opencv-4.9.0-150600.1.13 >  A f&ƢO ? /Ib5-OpcD<>K۰up$'`7 {wqFY.Qa90Mu)İCQ:FKL֋I/sl7L+ۨ?n‡w+TDWb=F -aO=L"xpWh0f~Q 2X,M~*d75%BMboW|IzـY5~Zɇ_^sm ?6}g/(ĘٍlGSg?MlwƖb,GĉXAkڲj!D\.q@zZ|ʯ.(z17_up6s3n-DB˜fXmѸ4\ ؎KZ6K'Mع Ÿ^ѝp#u ծd\*ZЭ=Lm}K:+;wUNh22`>w&wLㄒ/wijҎ!(n_{wa8096f51b0b6a55d4879eda443fbf535bfadd1ee30b219e35e3c666ccc0fa7b741d6861abf5e1c55cfa03f451caef6dabb1c9bffADf&ƢO ?VZLAqrxOW dK{|B]GQ[Y68`SlL-u]gbqR]1ȋ=Jr໷ wkωQ7 Ljl|`pа ߙJ(ā31hDXۀצ9OOj?;UZ_]AEb}\E3}J"Rf:7[ k<=&z'Gmv ܒTI) ZAZ&?؍جSڷ%8y/ک@Nf7XAogww##]jjfw"4zX6uRԃf7;U)7d5#ޚocpA?d   H 5V $ H l    z@d F( {8 @9 @:@BFGHIDXXYdZ[\]^?bc/defluv wLxyTz,<@FCopencv4.9.0150600.1.13Collection of algorithms for computer visionOpenCV means Intel Open Source Computer Vision Library. It is a collection of C functions and a few C++ classes that implement some popular Image Processing and Computer Vision algorithms.f&h01-ch3d&SUSE Linux Enterprise 15SUSE LLC BSD-3-Clause AND GPL-2.0-only AND Apache-2.0https://www.suse.com/Development/Libraries/C and C++https://opencv.org/linuxx86_64XH8i0,^,^,^D;A큤Af&f&f&f&f&f&eTf&eTeTeTeTeTeTeTeTeTf&ŝ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-4.9.0-150600.1.13.src.rpmopencvopencv(x86-64)opencv-qt5@@@@@@@@@@@@@@@@@@@@@@@@@    libc.so.6()(64bit)libc.so.6(GLIBC_2.14)(64bit)libc.so.6(GLIBC_2.2.5)(64bit)libc.so.6(GLIBC_2.3.4)(64bit)libc.so.6(GLIBC_2.34)(64bit)libc.so.6(GLIBC_2.4)(64bit)libgcc_s.so.1()(64bit)libgcc_s.so.1(GCC_3.0)(64bit)libm.so.6()(64bit)libm.so.6(GLIBC_2.2.5)(64bit)libopencv_calib3d.so.409()(64bit)libopencv_core.so.409()(64bit)libopencv_dnn.so.409()(64bit)libopencv_features2d.so.409()(64bit)libopencv_highgui.so.409()(64bit)libopencv_imgcodecs.so.409()(64bit)libopencv_imgproc.so.409()(64bit)libopencv_objdetect.so.409()(64bit)libopencv_videoio.so.409()(64bit)libstdc++.so.6()(64bit)libstdc++.so.6(CXXABI_1.3)(64bit)libstdc++.so.6(GLIBCXX_3.4)(64bit)libstdc++.so.6(GLIBCXX_3.4.11)(64bit)libstdc++.so.6(GLIBCXX_3.4.21)(64bit)libstdc++.so.6(GLIBCXX_3.4.9)(64bit)rpmlib(CompressedFileNames)rpmlib(FileDigests)rpmlib(PayloadFilesHavePrefix)rpmlib(PayloadIsXz)3.0.4-14.6.0-14.0-15.2-14.14.3ed@e@e_@c@c@bbk@bQu@aaa)@a,@`B`%@_`@__]@_^ _>e^^n@^&^%@^O@]@]@]fl]e@]Z@]N]M`@]G@[@[ @[ @ZkZS]@Z[@Z @ZYYZ@YYdYjY5GY5GY*W@WV@WEV@V@V'@V@VяVVzV^@VBU(U@U@T,@axel.braun@gmx.decabelo@opensuse.orgstefan.bruens@rwth-aachen.destefan.bruens@rwth-aachen.dechristophe@krop.frstefan.bruens@rwth-aachen.destefan.bruens@rwth-aachen.destefan.bruens@rwth-aachen.dealarrosa@suse.comstefan.bruens@rwth-aachen.dedmueller@suse.comjengelh@inai.destefan.bruens@rwth-aachen.dealoisio@gmx.comstefan.bruens@rwth-aachen.destefan.bruens@rwth-aachen.destefan.bruens@rwth-aachen.dedmueller@suse.commliska@suse.czstefan.bruens@rwth-aachen.destefan.bruens@rwth-aachen.dedimstar@opensuse.orgstefan.bruens@rwth-aachen.destefan.bruens@rwth-aachen.destefan.bruens@rwth-aachen.destefan.bruens@rwth-aachen.demarius.kittler@suse.comjengelh@inai.dejbrielmaier@suse.dechristophe@krop.frchristophe@krop.frstefan.bruens@rwth-aachen.dechristophe@krop.frfabian@ritter-vogt.defvogt@suse.comschwab@suse.detchvatal@suse.comstefan.bruens@rwth-aachen.dempluskal@suse.comecsos@opensuse.orgecsos@opensuse.orgmpluskal@suse.commpluskal@suse.comkah0922@gmail.comstefan.bruens@rwth-aachen.detchvatal@suse.comtchvatal@suse.comstefan.bruens@rwth-aachen.deolaf@aepfle.dedimstar@opensuse.orgmartin.liska@suse.comtoddrme2178@gmail.comtoddrme2178@gmail.comtittiatcoke@gmail.comjoerg.lorenzen@ki.tng.dejoerg.lorenzen@ki.tng.deolaf@aepfle.dealarrosa@suse.comohering@suse.deolaf@aepfle.demlin@suse.comcoolo@suse.comcoolo@suse.comtittiatcoke@gmail.com- Add %{?sle15allpythons} and build python bindings for all configured versions (also for Tumbleweed)- update to 4.9.0, highlights below, for details check https://github.com/opencv/opencv/wiki/ChangeLog#version490 Highlights of this release: * Core Module: + Added cv::broadcast + Fixed several rounding issues on ARM platform + Added detection & dispatching of some modern NEON instructions (NEON_FP16, NEON_BF16) + Added optimization for LoongArch 128-bit vector, detection & dispatching of LoongArch * DNN module patches: + Experimental transformers support + ONNX Attention layer support + ONNX Einsum layer support + OpenVINO backend for INT8 models + ONNX Gather Elements layer + ONNX InstanceNorm layer + Better support of ONNX Expand layer with cv::broadcast + Improved DNN graph fusion with shared nodes and commutative operations + New fastGEMM implementation and several layers on top of it + Winograd fp16 optimizations on ARM + Tests and multiple fixes for Yolo family models support + New layers support and bug fixes in CUDA backend: GEMM, Gelu, Add + CANN backend: bug fix, support HardSwish, LayerNormalization and InstanceNormalization + LayerNormalization: support OpenVINO, OpenCL and CUDA backend. * G-API module: + Intel® OpenVINO™ DL inference backend: - Introduced "inferenence only" ("benchmark") mode in the OV2.0 backend. - Fixed model layout setting issue in the OV2.0 backend. - Fixed/relaxed various asserts in the OV2.0 backend. + Core and image processing functionality: - Fluid kernels were rewritten to new universal intrinsics. Thanks for this contribution! + Streaming and video functionality: - Introduced a QueueSource: an alternative way to manually push input frames to the G-API pipeline in the streaming mode. - Introduced VAS Object Tracker (OT) for the various video analytics scenarios. + Python bindings: - Exposed VAS OT in G-API Python bindings. + Other changes and fixes: - Updated ADE (the G-API's graph library) to the latest version. - Various code clean-ups and warning fixes. * Objdetect module: + Implemented own QR code decoder as replacement for QUIRC library + Bug fixes in QR code encoder version estimation + More accurate Aruco marker corner refinement with dynamic window + Fixed contour filtering in ArUco + QR code detection sample for Android + Multiple local bug fixes and documentation update for Aruco makers, Charuco boards and QR codes. * Video: + Google Summer of Code: added a new object tracking API TrackerVit for a vision transformer-based VitTrack. This work is done by LIU Pengyu. * VideoIO: + Videoio: Add raw encoded video stream encapsulation to cv::VideoWriter with CAP_FFMPEG + Fix GStreamer backend with manual pipelines. * Calibration module: + Multiple fixes and improvements chess board calibration rig detector. + calibrateCamera throws exception, if calibration system is underconstrained. + Fixed bug in findEssentialMat with USAC + Fixed out-of-image access in cv::cornerSubPix + Fixed crash in ap3p + Fixed stereoRectify image boundaries + Fixed "use after free" issue in essential_solver.cpp * Python Bindings: + Added type stub generation for missed types and manually wrapped types. + Added read-only flag handling for Numpy arrays. + Fixed exception handling and bindings for in module. + Improved error messages in Numpy array type handling. + Fixed constructors documentation in Python. * Platforms and hardware Support: + Experimental CUDA support as first class language in CMake + Added experimental support for Apple VisionOS platform + Add support Orbbec Gemini2 and Gemini2 XL camera + Fix fullscreen behavior on macOS * Other: + OpenCV Summer of Code: semi-automated refactoring across multiple pull requests by HAN Liutong made our CPU-optimized code compatible with SIMD with variable vector length (RISC-V RVV)- update to 4.8.1 * WebP security update for CVE-2023-4863 * Depthwise convolution 5x5 performance regression fix - update to 4.8.0, highlights below, for details check https://github.com/opencv/opencv/wiki/ChangeLog#version480 Highlights of this release: * DNN module patches: + TFLite models support, including int8 quantized models. + Enabled DNN module build without Protobuf dependency. + Improved layers => supported more models: - ONNX: Layer normalization, GELU and QLinearSoftmax. - Fixes in CANN backend: * support ONNX Split, Slice, Clip (Relu6) and Conv with auto_pad. * support ONNX Sub, PRelu, ConvTranspose. - Reduce Refactor for robustness and potential follow-up improvements. - Fixes for Segment Anything Model by Meta. - Fixes in nary element wise layer about broadcast: * Fixes in CPU. * and Fixes in CUDA backend. - Further increased DNN speed on ARM and X86 by improving convolution, covering 1D and 3D cases, supporting convolution+element-wise op fusion. - Added full FP16 computation branch on ARMv8 platform, 1.5x faster than FP32 (FP16 Winograd is still pending). - Vulkan backend refactor for better performance and robustness. It runs 4X faster than before. - Added API blobFromImageParam to build network inputs with pre-processings. - Modern OpenVINO support. * G-API module: + Intel® OpenVINO™ inference backend: - Streamlined preprocessing in OpenVINO Inference Engine (ie) API 1.0 backend. Note: this backend will be deprecated after OpenVINO removes the API 1.0 support in its subsequent releases. - Aligned OpenVINO IE API 1.0 backend with the latest OpenVINO 2023.0 (as some features were removed there). - Introduced a brand new OpenVINO API 2.0 backend. - Implemented the required inference operations for the OpenVINO API 2.0 backend. + Python bindings: - Exposed varions normalization options for ONNX RT backend in Python bindings. - Exposed Fluid kernels and kernel package manipulation functions (combine()) in Python. - Fixed issues in Stateful Python kernel state handling; also fixed various issues in Python tests. - Fixed issue with opaque kernel output information handling which broke Python custom kernels. + Samples: - Introduced a new Segmentation demo with desync() to enable slow-running networks in the real-time. - Updated stats calculation in the G-API-based pipeline modelling tool. + Other changes and fixes: - Fixed tolerance in Fluid resize tests to avoid issues on ARM. - Fluid backend: extended Merge3 kernel with more supported data types. - Fixed standalone mode compilation issues. * Objdetect module: + FaceDetectorYN upgrade for better performance, accuracy and facial landmarks support. + New QR code detection algorithm based on ArUco code. + Bar code detector and decoder moved from Contrib to main repository. + Introduced common API for all graphical codes like bar codes and QR codes. + Added flag for legacy pre-4.6.0 ChAruco boards support. + Multiple bug fixes and improvements in QR code detection and decoding pipelines. + Multiple bug fixes and improvements in ArUco based pipelines. * Calibration module: + USAC framework improvements. + Fixed stddev estimation in camera calibration pipelines. + Fixed incorrect pixel grid generation in icvGetRectangles that improves accuracy of getOptimalNewCameraMatrix, stereoRectify and some other calibration functions. Charuco board support in patterns generator, interactive calibration tool and calibration samples. * Image processing module: + Various fixes in line segments detector. + Fixed even input dimensions for INTER_NEAREST_EXACT in resize. + Optimise local cost computation in IntelligentScissorsMB::buildMap. + Keep inliers for linear remap with BORDER_TRANSPARENT + Fix distransform to work with large images. * Features2d module: + SIFT accuracy improvements. * Core module: + Added REDUCE_SUM2 option to cv::reduce. + Introduced cv::hasNonZero function. + Update IPP binaries update to version 20230330. + Improved RISC-V RVV vector extensions support. - Support RVV v0.11 intrinsics available in LLVM 16 and GCC 13 - Support build with T-Head RISC-V toolchain (RVV 0.7.1 and 1.0) + Several OpenCL vendor and version handling improvements. * Multimedia: + Added AVIF support through libavif. + Orbbec Femto Mega cameras support. + HEVC/H265 support in VideoWriter with MS Media Foundation backend. + Fixed FPS computation on some videos for FFmpeg backend. + Added support for VideoCapture CAP_PROP_AUTO_WB and CV_CAP_PROP_WHITE_BALANCE_BLUE_U for DShow backend. + Fixes OBS Virtual Camera capture. + CV_32S encoding support with tiff. * Python Bindings: + Python typing stubs. + Fix reference counting errors in registerNewType. + Fixed ChAruco and diamond boards detector bindings. + Added bindings to allow GpuMat and Stream objects to be initialized from memory initialized in other libraries + np.float16 support. + Python bindings for RotatedRect, CV_MAKETYPE, CV_8UC(n). * JavaScript bindings: + Added possibility for disabling inlining wasm in opencv.js + Extended JS bindings for Aruco, Charuco, QR codes and bar codes. * Other: + Several critical issue fixes in wechat_qrcode module (opencv_contrib)- update to 4.7.0, highlights below, for details check https://github.com/opencv/opencv/wiki/ChangeLog#version470 Highlights of this release: * DNN: + New ONNX layers: Scatter and ScatterND, Tile, ReduceL1, ReduceMin and more. + Signinficant performance optimization for convolutions. Winograd algoritm implementation. + Element-wise operation (add, sub, mul, div, ...): Broadcasting. + OpenVino 2022.1 support. + CANN backend support. * Algorithms: + ArUco markers and April tags support including ChAruco and diamond boards detection and calibration. + QR code detection and decoding quality imrovement. Alignment markers support. Benchmark for QR codes: link + Nanotrack v2 tracker based on neural networks. + Stackblur algoruthm implementation. * Multimedia: + FFmpeg 5.x support. + CUDA 12.0 support. Hardware accelerated video codecs support on NVIDIA platforms with modern Video Codec SDK (NVCUVID and NVENCODEAPI). + CV_16UC1 read/write video support with FFmpeg. + Orientation meta support on Mac with native media API. + New iterator-based API for multi-page image formats. + libSPNG support for PNG format. + SIMD acceleration for self-built libJPEG-Turbo + H264/H265 support on Android. Multiple fixes for video decoder, endcoder and camera memory layout. * G-API + Exposed all core APIs to Python, including stateful kernels. * Optimization: + New universal intrinsics backend for scalable vector instructions. The first scalable implementation for RISC-V RVV 1.0. + DNN module patches: - Improved layers / supported more models: * Scatter and ScatterND #22529, Tile #22809 * Fixes in Slice (support negative step #22898) * Support some reduce layers of ONNX #21601 - Added CANN backend support #22634. Link to the manual: https://github.com/opencv/opencv/wiki/Huawei-CANN-Backend. - Added bacthed NMS for multi-class object detection #22857 - Accelerating convolution, especially for ARM CPU. - Winograd's convolution optimization + And many other contributions: + Added n-dimensional flip to core #22898 + Add StackBlur for imgproc #20379 - Removed upstream opencv-ffmpeg5.patch- Add upstream change to fix include issue with FFmpeg 5: * opencv-ffmpeg5.patch- update to 4.6.0, highlights below, for details check https://github.com/opencv/opencv/wiki/ChangeLog#version460 * OpenCV project infrastructure migrating on GitHub Actions workflows for CI and release purposes * Added support for GCC 12, Clang 15 * Added support for FFmpeg 5.0 * DNN module patches: + Improved layers / activations / supported more models: - LSTM (+CUDA), resize (+ONNX13), Sign, Shrink, Reciprocal, depth2space, space2depth - fixes in Reduce, Slice, Expand + Disabled floating-point denormals processing #21521 + Changed layer names in ONNX importer to support "output" entities properly + Added TIM-VX NPU backend support: https://github.com/opencv/opencv/wiki/TIM-VX-Backend-For-Running-OpenCV-On-NPU + Added Softmax parameter to ClassificationModel + Added audio speech recognition sample (C++) #21458 + Intel® Inference Engine backend (OpenVINO): - added initial support for OpenVINO 2022.1 release - removed support of legacy API (dropped since 2020.3) * G-API module: + G-API framework: - Introduced a Grayscale image format support for cv::MediaFrame: #21511; - Enabeled .reshape() support in the CPU backend: #21669; - Fixed possible hang in streaming execution mode with constant inputs: #21567; - Introduced proper error/exception propagation in the asynchronous streaming execution mode: #21660; - Fixed new stream event handling: #21731. + Fluid backend: - Fixed horizontal pass in the Resize kernel, fixed Valgrind issues: #21144; - Extended Resize kernel with F32 version: #21678, added AVX: #21728. - Enabled dynamic dispatch for Split4 kernel: #21520; - Enabled dynamic dispatch for Merge3 kernel: #21529; - Added a SIMD version for DivC kernel: #21474; - Added a SIMD version for DivRC kernel: #21530; - Enabled dynamic dispatch for Add kernel: #21686; - Enabled dynamic dispatch for Sub kernel: #21746; - Added a SIMD version for ConvertTo kernel: #21777; - Fixed kernel matrix size for Sobel kernel: #21613. + Intel® OpenVINO™ inference backend: - Fixed NV12 format support for remote memory when OpenVINO remote context is used: #21424. - Implemented correct error handling in the backend: #21579. - Fixed ngraph warnings #21362. + OpenCV AI Kit backend: - Introduced a new backend to program OpenCV AI Kit boards via G-API. Currently the backend is in experimental state, but allows to build Camera+NN pipeline and supports heterogeneity (mixing with host-side code): #20785, #21504. + Media integration: - Enabled GPU inference with oneVPL and DirectX11 on Windows in Intel OpenVINO inference backend: #21232, #21618, #21658, #21687, [#21688]. Now GPU textures decoded by oneVPL decoder can be preprocessed and inferred on GPU with no extra host processing. - Enabled oneVPL support on Linux: #21883. - Extended GStreamer pipeline source with Grayscale image format support: #21560. + Python bindings: - Exposed GStreamer pipeline source in Python bindings: #20832. - Fixed Python bindings for CudaBufferPool, cudacodec and cudastereo modules in OpenCV Contrib. + Samples: - Introduced a pipeline modelling tool for cascaded model benchmarking: #21477, #21636, #21719. The tool supports a declarative YAML-based config to describe pipelines with simulated pre-/post-processing. The tool collects and reports latency and throughput information for the modelled pipeline. + Other changes and fixes: - Moved GKernelPackage into cv:: namespace by default, its cv::gapi:: alias remain for compatibility: #21318; - Moved Resize kernel from core to imgproc kernel packages for CPU, OpenCL, and Fluid backends: #21157. Also moved tests appropriately: #21475; - Avoided sporadic test failures in DivC: #21626; - Fixed 1D Mat handling in the framework: #21782; - Reduced the number of G-API generated accuracy tests: #21909. - Drop upstream patches: * 0001-highgui-Fix-unresolved-OpenGL-functions-for-Qt-backe.patch * videoio_initial_FFmpeg_5_0_support.patch * videoio_ffmpeg_avoid_memory_leaks.patch- Add upstream patches for FFmpeg 5.0 support, add * videoio_initial_FFmpeg_5_0_support.patch * videoio_ffmpeg_avoid_memory_leaks.patch- Restore memoryperjob constraint, avoid being scheduled on a 16 core system and use less than half of it. - Adjust %limit_build to 1800, to avoid recurrent build failures on aarch64. (People should not care for their pet architecture only, but also carefully check if they break others.) - Add missing libopencv_aruco dependency in devel package.- Remove the memoryperjob constraint which doesn't work as one would expect and breaks ppc64 builds. - Use %limit_memory -m 1700 to set the number of concurrent jobs to a sane value and fix OOM errors when building in workers with many cores. - Decrease the disk constraint to 9G which seems to be enough- update to 4.5.5, highlights below, for details check https://github.com/opencv/opencv/wiki/ChangeLog#version455 * Audio support as part of VideoCapture API: GStreamer #21264 * Updated SOVERSION handling rules: #21178 * DNN module patches: + Added tests to cover ONNX conformance test suite: #21088 + Improved layers / activations / supported more models + Upgraded builtin protobuf from 3.5.2 to 3.19.1 + More optimizations for RISC-V platform + Intel® Inference Engine backend ( OpenVINO™ ): added support for OpenVINO 2021.4.2 LTS release * G-API module: + G-API framework: - Fixed issue with accessing 1D data from cv::RMat: #21103 - Restricted passing the G-API types to graph inputs/outputs for execution: #21041 - Various fixes in G-API Doxygen reference: #20924 - Renamed various internal structures for consistency #20836 #21040 + Fluid backend: - Introduced a better vectorized version of Resize: #20664. - Added vectorized version of Multiply kernel: #21024 - Added vectorized version of Divide kernel: #20914 - Added vectorized version of AddC kernel: #21119 - Added vectorized version of SubC kernel: #21158 - Added vectorized version of MulC kernel: #21177 - Added vectorized version of SubRC kernel: #21231 - Enabled SIMD dispatching for AbsDiffC: #21204 + OpenCL backend: - Fixed sporadic test failures in Multiply kernel running on GPU: #21205 + Intel® OpenVINO™ inference backend: - Extended ie::Params to support static batch size as input to inference: #20856 - Enabled 2D input tensor support in IE backend: #20925 - Fixed various issues with imported (pre-compiled) networks: #20918 + Media integration: - Introduced a GStreamer-based pipeline source for G-API: #20709 - Completed the integration of Intel® oneVPL as a pipeline source for G-API #20773 with device selection #20738, asynchronous execution #20901, intial demux support #21022, and GPU-side memory allocation via DirectX 11 #21049. + Samples: - Replaced custom kernels with now-standard G-API operations in several samples #21106 - Moved API snippets from G-API samples to a dedicated place #20857 + Other changes and fixes: - Fixed various static analysis issues for OpenVINO 2021.4 release: #21083 and #21212 - Fixed various build warnings introduced after OpenVINO update: #20937 - Continued clean-up in the G-API test suite on GTest macros [#20922] and test data #20995 - Added custom accuracy comparison functions to Fluid performance tests: #21150. * And many other contributions: + Added QRcode encoder: #17889 + GSoC - OpenCV.js: Accelerate OpenCV.js DNN via WebNN: #20406 + Add conventional Bayer naming: #20970 + (opencv_contrib) Add Radon transform function to ximgproc: #3090 + (opencv_contrib) New superpixel algorithm (F-DBSCAN): #3093 + Created Stitching Tool: #21020 + Improve CCL with new algorithms and tests: #21275 + (opencv_contrib) Update ArUco tutorial: #3126 - Adjust memory constraints (mostly required for aarch64 on Leap) - Add 0001-highgui-Fix-unresolved-OpenGL-functions-for-Qt-backe.patch- update to 4.5.4: * 8-bit quantization in the dnn module * Improved Julia bindings * Speech recognition sample * dnn module optimizations for RISC-V * Tutorial about universal intrinsics and parallel_for usage * Improvements in the dnn module: - New layers and models support - Some existing layers have been fixed - Soft-NMS implementation - Supported OpenVINO 2021.4.1 LTS release- Remove dependency on IlmBase, opencv never uses this directly.- update to 4.5.2, highlights below, for details check https://github.com/opencv/opencv/wiki/ChangeLog#version452 * core: added support for parallel backends. * imgproc: added IntelligentScissors implementation (JS demo). * videoio: improved hardware-accelerated video de-/encoding tasks. * DNN module: + Improved debugging of TensorFlow parsing errors: #19220 + Improved layers / activations / supported more models: - optimized: NMS processing, DetectionOutput - fixed: Div with constant, MatMul, Reshape (TensorFlow behaviour) - added support: Mish ONNX subgraph, NormalizeL2 (ONNX), LeakyReLU (TensorFlow), TanH + SAM (Darknet), Exp + Intel® Inference Engine backend ( OpenVINO™ ): added support for OpenVINO 2021.3 release * G-API module: + Python support: - Introduced a new Python backend - now G-API can run custom kernels written in Python as part of the pipeline: #19351 - Extended Inference support in the G-API bindings: #19318 - Added more graph data types in the G-API bindings: #19319 + Inference support: - Introduced dynamic input / CNN reshape functionality in the OpenVINO inference backend #18240 - Introduced asynchronous execution support in the OpenVINO inference backend, now it can run in multiple parallel requests to increase stream density/throughput: #19487, #19425 - Extended supported data types with INT64/INT32 in ONNX inference backend and with INT32 in the OpenVINO inference backend #19792 - Introduced cv::GFrame / cv::MediaFrame and constant support in the ONNX backend: #19070 + Media support: - Introduced cv::GFrame / cv::MediaFrame support in the drawing/rendering interface: #19516 - Introduced multi-stream input support in Streaming mode and frame synchronization policies to support cases like Stereo: #19731 - Added Y and UV operations to access NV12 data of cv::GFrame at the graph level; conversions are done on-the-fly if the media format is different: #19325 + Operations and kernels: - Added performance tests for new operations (MorphologyEx, BoundingRect, FitLine, FindContours, KMeans, Kalman, BackgroundSubtractor) - Fixed RMat input support in the PlaidML backend: #19782 - Added ARM NEON optimizations for Fluid AbsDiffC, AddWeighted, and bitwise operations: #18466, #19233 - Other various static analysis and warning fixes + Documentation: - [GSoC] Added TF/PyTorch classification conversion: #17604 - [GSoC] Added TF/PyTorch segmentation conversion: #17801 - [GSoC] Added TF/PyTorch detection model conversion: #18237 - Updated documentation to address Wide Universal Intrinsics (WUI) SIMD API: #18952 + And many other great contributions from OpenCV community: - core: cuda::Stream constructor with stream flags: #19286 - highgui: pollKey() implementation for w32 backend: #19411 - imgcodecs: Added Exif parsing for PNG: #19439 - imgcodecs: OpenEXR compression options: #19540 - imgproc: connectedComponents optimizations: (Spaghetti Labeling): #19631 - videoio: Android NDK camera support #19597 - (contrib) WeChat QRCode module open source: #2821 - (contrib) Implemented cv::cuda::inRange(): #2803 - (contrib) Added algorithms from Edge Drawing Library: #2313 - (contrib) Added Python bindings for Viz module: #2882 - Add libva build dependency for HW accelerated videoio - Slight bump for memory constraints- Enable aruco module (recognize markers to detect camera pose)- update to 4.5.1, highlights below, for details check https://github.com/opencv/opencv/wiki/ChangeLog#version451 * Continued merging of GSoC 2020 results: + Develop OpenCV.js DNN modules for promising web use cases together with their tutorials + OpenCV.js: WASM SIMD optimization 2.0 + High Level API and Samples for Scene Text Detection and Recognition + SIFT: SIMD optimization of GaussianBlur 16U * DNN module: + Improved layers / activations / supported more models: - optimized: 1D convolution, 1D pool - fixed: Resize, ReduceMean, Gather with multiple outputs, importing of Faster RCNN ONNX model - added support: INT32 ONNX tensors + Intel® Inference Engine backend (OpenVINO): - added support for OpenVINO 2021.2 release - added preview support for HDDL + Fixes and optimizations in DNN CUDA backend (thanks to @YashasSamaga) * G-API Framework: + Introduced serialization for cv::RMat, including serialization for user-defined memory adapters + Introduced desync, a new Operation for in-graph asynchronous execution - to allow different parts of the graph run with a different latency + Introduced a notion of "in-graph metadata", now various media-related information can be accessed in graph directly (currently only limited to timestamps and frame IDs) + Introduced a new generic task-based executor, based on Threading Building Blocks (TBB) + Extended infer<>() API to accept a new cv::GFrame data structure to allow handling of various media formats without changes in the graph structure + Made copy() an intrinsic where real copy may not happen (optimized out) based on graph structure, extended it to support cv::GFrame + Various fixes, including addressig static analysis, documentation, and test issues * G-API Operations: + Introduced new operations morphologyEx, boundingRect, fitLine, kmeans, Background Subtractor, Kalman filter * G-API Intel® Inference Engine backend (OpenVINO): + Extended cv::gapi::ie::Params<> to import CNN networks (e.g. pre-compiled ones) instead of passing .XML and .BIN files; also enabled configuring Inference Engine plugins via this structure + Added a new overload to infer<>() to run inference over a single region of interest + Added support for cv::MediaFrame input data type (projected from cv::GFrame) and handling for NV12 input image format * G-API Python bindings: + Exposed G-API's Inference and Streaming APIs in the OpenCV Python bindings + Added initial Python support for cv::GArray data structure * Significant progress on RISC-V port. - Updated constraints, bump memory to 5 GB - Cleaned up spec file- Split library package, move all libraries with external dependencies (Qt5, ffmpeg, gstreamer) into separate packages - Move haar and LBP cascades into separate package, pull in from objdetect and face (detect) libraries.- update to 4.5.0, see https://github.com/opencv/opencv/wiki/ChangeLog#version450 for details, highlights: * OpenCV license has been changed to Apache 2 (OpenCV 3.x will keep using BSD) * GSoC is over, all projects were success and most of them have already been merged. Optimizations for RISC-V, bindings for Julia language, real-time single object tracking, improved SIFT and others * OpenJPEG is now used by default for JPEG2000 * Supported multiple OpenCL contexts * Improvements in dnn module: + Support latest OpenVINO 2021.1 release + Tengine lite support for inference on ARM + Many fixes and optimizations in CUDA backend * Added Python bindings for G-API module * Multiple fixes and improvements in flann module * Added Robot-World/Hand-Eye calibration function- update to 4.4.0: * SIFT (Scale-Invariant Feature Transform) algorithm has been moved to the main repository (patent on SIFT is expired) * DNN module: * State-of-art Yolo v4 Detector: #17148. * onnx: Add support for Resnet_backbone * EfficientDet models * add text recognition sample / demo * FlowNet2 optical flow * Intel Inference Engine backend * added support for OpenVINO 2020.3 LTS / 2020.4 releases * support of NN Builder API is planned for removal in the next release * Many fixes and optimizations in CUDA backend * Obj-C / Swift bindings: #17165 * Julia bindings as part of ongoing GSoC project * BIMEF: A Bio-Inspired Multi-Exposure Fusion Framework for Low-light Image Enhancement * Enable Otsu thresholding for CV_16UC1 images * Add Stroke Width Transform algorithm for Text Detection * Planned migration on Apache 2 license for next releases - remove opencv-includedir.patch (obsolete)- Use memoryperjob constraint instead of %limit_build macro.- Update to 4.3.0 * DNN module: + Improved layers / activations / supported more models: - ONNX: LSTM, Broadcasting, Algebra over constants, Slice with multiple inputs - DarkNet: grouped convolutions, sigmoid, swish, scale_channels - MobileNet-SSD v3: #16760 + New samples / demos: - Clothes parts segmentation and CP-VTON - DaSiamRPN tracker Intel® Inference Engine backend (OpenVINO™): - added support for custom layers through nGraph OpenVINO API: #16628 - nGraph OpenVINO API is used by default: #16746 + Many fixes and optimizations in CUDA backend (thanks to @YashasSamaga) + OPEN AI LAB team submitted the patch that accelerates OpenCV DNN on ARM using their Tengine library * G-API module: + Introduced a new graph-level data type GOpaque. This type can be used to pass arbitrary user data types between G-API nodes in the graph (supported for CPU/OpenCV backend only). + Introduced a way to declare G-API CPU (OpenCV) kernels in-place + Added a new sample "Privacy masking camera", combining Deep Learning with traditional Image Processing (link) + Added more operations in the default library: WarpAffine, WarpPerspective, NV12toGray. * Performance improvements: + IPP-ICV library with CPU optimizations has been updated to version 2020.0.0 Gold + SIMD intrinsics: integral, resize, (opencv_contrib) RLOF implementation #2476 * And many other great contributions from OpenCV community: + (opencv_contrib) Computer Vision based Alpha Matting (GSoC 2019) #2306 + calib3d: findChessboardCornersSB improvements: #16625 + calib3d: updated documentation for RT matrices: #16860 + core: improved getNumberOfCPUs(): #16268 + imgproc: new algorithm HOUGH_GRADIENT_ALT is added to HoughCircles() function #16561. It has much better recall and precision + imgcodecs: added initial support for OpenJPEG library (version 2+): #16494 + highgui(Qt): added Copy to clipboard: #16677 + dnn: TensorFlow, Darknet and ONNX importers improvements by @ashishkrshrivastava + (opencv_contrib) added rapid module for silhouette based 3D object tracking: #2356 + (opencv_contrib) SIFT detector is enabled by default due patents expiration (without requirement of NONFREE build option) + help materials: OpenCV Cheat Sheet in Python: #4875 * Changes that can potentially break compatibility: + image filtering functions throws exception on empty input (voting results) - Packaging changes: * Stop mangling CMake diagnostic output, no dependency versions end up in the packages anyway, drop opencv-build-compare.patch * Set empty OPENCV_DOWNLOAD_TRIES_LIST, skip downloads even when network is available during builds (e.g. local build). * Drop upstream GLES patches: + 0001-Do-not-include-glx.h-when-using-GLES.patch + opencv-gles.patch- Disable Python 2 bindings for Tumbleweed.- Drop Jasper (i.e jpeg2k) support (boo#1130404, boo#1144260) JasPer is unmaintained, CVEs are not being addressed (some issues received patches submitted to the upstream github project, but are not being merged, other CVEs are considered unfixable). openSUSE follows other distros in dropping JasPer now (much later than most others, incl. Debian).- Add webp build dependency to use system libwebp instead of bundled one. - Enable dispatch of AVX512 optimized code.- Update to 4.2.0 * DNN module: + Integrated GSoC project with CUDA backend: #14827 + Intel® Inference Engine backend ( OpenVINO™ ): - support for nGraph OpenVINO API (preview / experimental): #15537 * G-API module: + Enabled in-graph inference: #15090. Now G-API can express more complex hybrid CV/DL algorithms; - Intel® Inference Engine backend is the only available now, support for DNN module will be added in the future releases. + Extended execution model with streaming support: #15216. Decoding, image processing, inference, and post-processing are now pipelined efficiently when processing a video stream with G-API. + Added tutorials covering these new features: Face analytics pipeline and a sample Face beautification algorithm. * Performance improvements: + SIMD intrinsics: StereoBM/StereoSGBM algorithms, resize, integral, flip, accumulate with mask, HOG, demosaic, moments + Muti-threading: pyrDown * And many other great patches from OpenCV community: + VideoCapture: video stream extraction (demuxing) through FFmpeg backend. + VideoCapture: waitAny() API for camera input multiplexing (Video4Linux through poll() calls). + (opencv_contrib) new algorithm Rapid Frequency Selective Reconstruction (FSR): #2296 + tutorial. + (opencv_contrib) RIC method for sparse match interpolation: #2367. + (opencv_contrib) LOGOS features matching strategy: #2383. * Breaking changes: + Disabled constructors for legacy C API structures. + Implementation of Thread Local Storage (TLS) has been improved to release data from terminated threads. API has been changed. + Don't define unsafe CV_XADD implementation by default. + Python conversion rules of passed arguments will be updated in next releases: #15915.- Limit build parallelism with limit_build, some ARM and PPC workers have a high SMP/memory ratio and run out of memory otherwise. - Apply memory constraints (3GB) to all architectures, avoid being scheduled on very weak workers.- Update to 4.1.2 * DNN module: + Intel Inference Engine backend (OpenVINO): - 2019R3 has been supported - Support modern IE Core API - New approach for custom layers management. Now all the OpenCV layers fallbacks are implemented as IE custom layers which helps to improve efficiency due less graph partitioning. - High-level API which introduces dnn::Model class and set of task-specific classes such dnn::ClassificationModel, dnn::DetectionModel, dnn::SegmentationModel. It supports automatic pre- and post-processing for deep learning networks. * Performance improvements and platforms support: + MSA SIMD implementation has been contributed for MIPS platforms: https://github.com/opencv/opencv/pull/15422 + OpenCV.js optimization (threading and SIMD as part of GSoC project): https://github.com/opencv/opencv/pull/15371 + More optimizations using SIMD intrinsics: dotProd, FAST corners, HOG, LK pyramid (VSX), norm, warpPerspective, etc + Fixed detection of Cascade Lake CPUs * And many other great patches from OpenCV community: + GUI: support topmost window mode (Win32/COCOA): https://github.com/opencv/opencv/pull/14872 + Java: fix Mat.toString() for higher dimensions: https://github.com/opencv/opencv/pull/15181 + Implementation of colormap "Turbo" https://github.com/opencv/opencv/pull/15388 + QR-Code detection accuracy improvement: https://github.com/opencv/opencv/pull/15356 + GSoC: Add learning-based super-resolution module: https://github.com/opencv/opencv_contrib/pull/2229 and https://github.com/opencv/opencv_contrib/pull/2231 + Detection accuracy improvement of the white marker aruco corners: https://github.com/opencv/opencv_contrib/pull/2236 + Added pattern generator tool for aruco: https://github.com/opencv/opencv_contrib/pull/2250 + and special thanks to @sturkmen72 for improvind and cleaning up code of samples/tutorials * Breaking changes: + fixed values thresholding accuracy in calcHist() * Security fixes: CVE-2019-15939 (boo#1149742). - Enable Graph API (G-API) - Minor spec file cleanup- Include pkg-config file in opencv-devel package * Add opencv-includedir.patch- Avoid use of ®/™ signs in specfiles as per guidelines.- Disable LTO on ppc64le for now, as it fails to build when enabled (boo#1146096).- Increase the disk space needed to build opencv.- Update to 4.1.1 * DNN module: * 3D convolution networks initial support * A lot of improvements for ONNX and TenforFlow importers * Performance improvements * Added IPPE method for planar pose estimation in solvePnP * Added solvePnPRefineLM and solvePnPRefineVVS * Security fixes: CVE-2019-14491 (boo#1144352), CVE-2019-14492 (boo#1144348). - Check https://github.com/opencv/opencv/wiki/ChangeLog#version411 for the complete list of changes. - Drop fix_processor_detection_for_32bit_on_64bit.patch. Fixed upstream - Drop 0001-Handle-absolute-OPENCV_INCLUDE_INSTALL_PATH-correctl.patch Fixed upstream - Refresh 0001-Do-not-include-glx.h-when-using-GLES.patch and opencv-build-compare.patch.- Update to version 4.1.0 * DNN module: + Reduced peak memory consumption for some models up to 30%. + Inference Engine - Inference Engine 2018R3 is now a minimal supported version of IE. - Myriad X (Intel® Neural Compute Stick 2) is now supported and tested. - Automatic IR network reshaping for different inputs. - Improved samples to work with models from OpenVINO Open Model Zoo + New networks from TensorFlow Object Detection API: Faster-RCNNs, SSDs and Mask-RCNN with dilated convolutions, FPN SSD * Performance improvements: + More optimization using AVX2 instruction set. + Automatic runtime dispatching is available for large set of functions from core and imgproc modules. * Other improvements: + Matplotlib Perceptually Uniform Sequential colormaps + Add keypoints matching visualization for real-time pose estimation tutorial + Add Hand-Eye calibration methods + Java: improved support for multidimensional arrays (Mat) + Dynamically loaded videoio backends (FFmpeg, GStreamer) + opencv_contrib: Robust local optical flow (RLOF) implementations + opencv_contrib: Implementation of Quasi Dense Stereo algorithm + opencv_contrib: New module: Image Quality Analysis (IQA) API + opencv_contrib: BRISQUE No Reference Image Quality Assessment (IQA) API Check https://github.com/opencv/opencv/wiki/ChangeLog#version410 - Update to version 4.0.0 * A lot of C API from OpenCV 1.x has been removed. The affected modules are objdetect, photo, video, videoio, imgcodecs, calib3d. * Persistence (storing and loading structured data to/from XML, YAML or JSON) in the core module has been completely reimplemented. * OpenCV is now C++11 library and requires C++11-compliant compiler. Thanks to the extended C++11 standard library, we could get rid of hand-crafted cv::String and cv::Ptr. Now cv::String == std::string and cv::Ptr is a thin wrapper on top of std::shared_ptr. Also, on Linux/BSD for cv::parallel_for_ we now use std::thread's instead of pthreads. * DNN improvements * Completely new module opencv_gapi has been added. It is the engine for very efficient image processing, based on lazy evaluation and on-fly construction. * Performance improvements A few hundreds of basic kernels in OpenCV have been rewritten using so-called "wide universal intrinsics". Those intrinsics map to SSE2, SSE4, AVX2, NEON or VSX intrinsics, depending on the target platform and the compile flags. * QR code detector and decoder have been added to opencv/objdetect module. * The popular Kinect Fusion algorithm has been implemented, optimized for CPU and GPU (OpenCL), and integrated into opencv_contrib/rgbd module. * Very efficient and yet high-quality DIS dense optical flow algorithm has been moved from opencv_contrib to opencv, video module. See the example. * The slower TV L1 optical flow algorithm has been moved to opencv_contrib. Check https://github.com/opencv/opencv/wiki/ChangeLog#version400 - Drop obsolete opencv-lib_suffix.patch - Add 0001-Handle-absolute-OPENCV_INCLUDE_INSTALL_PATH-correctl.patch - As this is a major version upgrade, the old 3.4.x package is still available as opencv3- Update to 3.4.3 * Compatibility fixes with python 3.7 * Added a new computational target DNN_TARGET_OPENCL_FP16 * Extended support of Intel's Inference Engine backend * Enabled import of Intel's OpenVINO pre-trained networks from intermediate representation (IR). * tutorials improvements Check https://github.com/opencv/opencv/wiki/ChangeLog#version343 for the complete changelog. - Drop fix-build-i386-nosse.patch, build-workaround-issues-with-c.patch (fixed upstream) - Refresh patches- Add patch to fix use of headers from C: * build-workaround-issues-with-c.patch- Update to 3.4.1: * Added support for quantized TensorFlow networks * OpenCV is now able to use Intel DL inference engine as DNN acceleration backend * Added AVX-512 acceleration to the performance-critical kernels * Fix cmake mapping of RelWithDebInfo (boo#1154091). * For more information, read https://github.com/opencv/opencv/wiki/ChangeLog#version341 - Update contrib modules to 3.4.1: * No changelog available - Change mechanism the contrib modules are built - Include LICENSE of contrib tarball as well - Build with python3 on >= 15 - Add patch to fix build on i386 without SSE: * fix-build-i386-nosse.patch - Refresh patches: * fix_processor_detection_for_32bit_on_64bit.patch * opencv-build-compare.patch - Mention all libs explicitly - Rebase 3.4.0 update from i@marguerite.su - update to 3.4.0 * Added faster R-CNN support * Javascript bindings have been extended to cover DNN module * DNN has been further accelerated for iGPU using OpenCL * On-disk caching of precompiled OpenCL kernels has been finally implemented * possible to load and run pre-compiled OpenCL kernels via T-API * Bit-exact 8-bit and 16-bit resize has been implemented (currently supported only bilinear interpolation) - update face module to 3.4.0 - add opencv-lib_suffix.patch, remove LIB_SUFFIX from OPENCV_LIB_INSTALL_PATH, as CMAKE_INSTALL _LIBDIR is arch dependent.- Add option to build without openblas- Add conditionals for python2 and python3 to allow us enabling only desired python variants when needed - Do not depend on sphinx as py2 and py3 seem to collide there- Readd opencv-gles.patch, it is *not* included upstream; otherwise build breaks on all GLES Qt5 platforms (armv6l, armv7l, aarch64) - add fix_processor_detection_for_32bit_on_64bit.patch - Correctly set optimizations and dynamic dispatch on ARM, use OpenCV 3.3 syntax on x86.- Update licensing information- change requires of python-numpy-devel to build in Leap and to not break factory in future- fix build error/unresolvable for Leap 42.2 and 42.3- Update to version 3.3.1: * Lots of various bugfixes - Update source url- Rename python subpackage to python2 - Do not explicitly require python-base for python subpackages- Update to 3.3 - Dropped obsolete patches * opencv-gcc6-fix-pch-support-PR8345.patch * opencv-gles.patch - Updated opencv-build-compare.patch- Add 0001-Do-not-include-glx.h-when-using-GLES.patch Fix build for 32bit ARM, including both GLES and desktop GL headers causes incompatible pointer type errors- Add conditional for the qt5/qt4 integration * This is used only for gui tools, library is not affected - Add provides/obsoletes for the qt5 packages to allow migration - Drop patch opencv-qt5-sobump.diff * Used only by the obsoleted qt5 variant- Cleanup a bit with spec-cleaner - Use %cmake macros - Remove the conditions that are not really needed - Add tests conditional disabled by default * Many tests fail and there are missing testdata - Switch to pkgconfig style dependencies- Update to OpenCV 3.2.0 - Results from 11 GSoC 2016 projects have been submitted to the library: + sinusoidal patterns for structured light and phase unwrapping module [Ambroise Moreau (Delia Passalacqua)] + DIS optical flow (excellent dense optical flow algorithm that is both significantly better and significantly faster than Farneback’s algorithm – our baseline), and learning-based color constancy algorithms implementation [Alexander Bokov (Maksim Shabunin)] + CNN based tracking algorithm (GOTURN) [Tyan Vladimir (Antonella Cascitelli)] + PCAFlow and Global Patch Collider algorithms implementation [Vladislav Samsonov (Ethan Rublee)] + Multi-language OpenCV Tutorials in Python, C++ and Java [João Cartucho (Vincent Rabaud)] + New camera model and parallel processing for stitching pipeline [Jiri Horner (Bo Li)] + Optimizations and improvements of dnn module [Vitaliy Lyudvichenko (Anatoly Baksheev)] + Base64 and JSON support for file storage. Use names like “myfilestorage.xml?base64” when writing file storage to store big chunks of numerical data in base64-encoded form. [Iric Wu (Vadim Pisarevsky)] + tiny_dnn improvements and integration [Edgar Riba (Manuele Tamburrano, Stefano Fabri)] + Quantization and semantic saliency detection with tiny_dnn [Yida Wang (Manuele Tamburrano, Stefano Fabri)] + Word-spotting CNN based algorithm [Anguelos Nicolaou (Lluis Gomez)] - Contributions besides GSoC: + Greatly improved and accelerated dnn module in opencv_contrib: - Many new layers, including deconvolution, LSTM etc. - Support for semantic segmentation and SSD networks with samples. - TensorFlow importer + sample that runs Inception net by Google. + More image formats and camera backends supported + Interactive camera calibration app + Multiple algorithms implemented in opencv_contrib + Supported latest OSes, including Ubuntu 16.04 LTS and OSX 10.12 + Lot’s of optimizations for IA and ARM archs using parallelism, vector instructions and new OpenCL kernels. + OpenCV now can use vendor-provided OpenVX and LAPACK/BLAS (including Intel MKL, Apple’s Accelerate, OpenBLAS and Atlas) for acceleration - Refreshed opencv-build-compare.patch - Dropped upstream opencv-gcc5.patch - Replace opencv-gcc6-disable-pch.patch with upstream patch opencv-gcc6-fix-pch-support-PR8345.patch - Enable TBB support (C++ threading library) - Add dependency on openBLAS- Enable ffmpeg support unconditional- In case we build using GCC6 (or newer), add -mlra to CFLAGS to workaround gcc bug https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71294.- Apply upstream patch opencv-gcc6-disable-pch.patch to disable PCH for GCC6.- Test for python versions greater than or equal to the current version.- Add python 3 support- Added opencv_contrib_face-3.1.0.tar.bz2 * This tarball is created to take only the face module from the contrib package. The Face module is required by libkface, which in its turn is required by digikam.- Added _constraints file to avoid random failures on small workers (at least for builds on PMBS)- Update to OpenCV 3.1.0 - A lot of new functionality has been introduced during Google Summer of Code 2015: + “Omnidirectional Cameras Calibration and Stereo 3D Reconstruction” – opencv_contrib/ccalib module (Baisheng Lai, Bo Li) + “Structure From Motion” – opencv_contrib/sfm module (Edgar Riba, Vincent Rabaud) + “Improved Deformable Part-based Models” – opencv_contrib/dpm module (Jiaolong Xu, Bence Magyar) + “Real-time Multi-object Tracking using Kernelized Correlation Filter” – opencv_contrib/tracking module (Laksono Kurnianggoro, Fernando J. Iglesias Garcia) + “Improved and expanded Scene Text Detection” – opencv_contrib/text module (Lluis Gomez, Vadim Pisarevsky) + “Stereo correspondence improvements” – opencv_contrib/stereo module (Mircea Paul Muresan, Sergei Nosov) + “Structured-Light System Calibration” – opencv_contrib/structured_light (Roberta Ravanelli, Delia Passalacqua, Stefano Fabri, Claudia Rapuano) + “Chessboard+ArUco for camera calibration” – opencv_contrib/aruco (Sergio Garrido, Prasanna, Gary Bradski) + “Implementation of universal interface for deep neural network frameworks” – opencv_contrib/dnn module (Vitaliy Lyudvichenko, Anatoly Baksheev) + “Recent advances in edge-aware filtering, improved SGBM stereo algorithm” – opencv/calib3d and opencv_contrib/ximgproc (Alexander Bokov, Maksim Shabunin) + “Improved ICF detector, waldboost implementation” – opencv_contrib/xobjdetect (Vlad Shakhuro, Alexander Bovyrin) + “Multi-target TLD tracking” – opencv_contrib/tracking module (Vladimir Tyan, Antonella Cascitelli) + “3D pose estimation using CNNs” – opencv_contrib/cnn_3dobj (Yida Wang, Manuele Tamburrano, Stefano Fabri) - Many great contributions made by the community, such as: + Support for HDF5 format + New/Improved optical flow algorithms + Multiple new image processing algorithms for filtering, segmentation and feature detection + Superpixel segmentation and much more - IPPICV is now based on IPP 9.0.1, which should make OpenCV even faster on modern Intel chips - opencv_contrib modules can now be included into the opencv2.framework for iOS - Newest operating systems are supported: Windows 10 and OSX 10.11 (Visual Studio 2015 and XCode 7.1.1) - Interoperability between T-API and OpenCL, OpenGL, DirectX and Video Acceleration API on Linux, as well as Android 5 camera. - HAL (Hardware Acceleration Layer) module functionality has been moved into corresponding basic modules; the HAL replacement mechanism has been implemented along with the examples - Removed improve-sphinx-search.diff, opencv-altivec-vector.patch, opencv-pkgconfig.patch and opencv-samples.patch, fixed upstream. - Fixed opencv-qt5-sobump.diff, opencv-build-compare.patch, opencv-gcc5.patch and opencv-gles.patch. - Version OpenCV 3.0.0 + ~1500 patches, submitted as PR @ github. All our patches go the same route. + opencv_contrib (http://github.com/itseez/opencv_contrib) repository has been added. A lot of new functionality is there already! opencv_contrib is only compatible with 3.0/master, not 2.4. Clone the repository and use “cmake … - D OPENCV_EXTRA_MODULES_PATH= …” to build opencv and opencv_contrib together. + a subset of Intel IPP (IPPCV) is given to us and our users free of charge, free of licensing fees, for commercial and non-commerical use. It’s used by default in x86 and x64 builds on Windows, Linux and Mac. + T-API (transparent API) has been introduced, this is transparent GPU acceleration layer using OpenCL. It does not add any compile-time or runtime dependency of OpenCL. When OpenCL is available, it’s detected and used, but it can be disabled at compile time or at runtime. It covers ~100 OpenCV functions. This work has been done by contract and with generous support from AMD and Intel companies. + ~40 OpenCV functions have been accelerated using NEON intrinsics and because these are mostly basic functions, some higher-level functions got accelerated as well. + There is also new OpenCV HAL layer that will simplifies creation of NEON-optimized code and that should form a base for the open-source and proprietary OpenCV accelerators. + The documentation is now in Doxygen: http://docs.opencv.org/master/ + We cleaned up API of many high-level algorithms from features2d, calib3d, objdetect etc. They now follow the uniform “abstract interface – hidden implementation” pattern and make extensive use of smart pointers (Ptr<>). + Greatly improved and extended Python & Java bindings (also, see below on the Python bindings), newly introduced Matlab bindings (still in alpha stage). + Improved Android support – now OpenCV Manager is in Java and supports both 2.4 and 3.0. + Greatly improved WinRT support, including video capturing and multi-threading capabilities. Thanks for Microsoft team for this! + Big thanks to Google who funded several successive GSoC programs and let OpenCV in. The results of many successful GSoC 2013 and 2014 projects have been integrated in opencv 3.0 and opencv_contrib (earlier results are also available in OpenCV 2.4.x). We can name: - text detection - many computational photography algorithms (HDR, inpainting, edge-aware filters, superpixels, …) - tracking and optical flow algorithms - new features, including line descriptors, KAZE/AKAZE - general use optimization (hill climbing, linear programming) - greatly improved Python support, including Python 3.0 support, many new tutorials & samples on how to use OpenCV with Python. - 2d shape matching module and 3d surface matching module - RGB-D module - VTK-based 3D visualization module - etc. + Besides Google, we enjoyed (and hope that you will enjoy too) many useful contributions from community, like: - biologically inspired vision module - DAISY features, LATCH descriptor, improved BRIEF - image registration module - etc.- Reduce build-compare noise opencv-build-compare.patch- Remove BuildRequirement for python-sphinx in SLE12, since it's not available there and it's not a mandatory requirement.- Reduce differences between two spec files- Use pkgconfig for ffmpeg BuildRequires- Update improve-sphinx-search.diff for new python-Sphinx(1.3.1) * now that sphinx-build disallow executing without arguments and give you "Insufficient arguments" error, use "sphinx-build -h" instead * the default usages output ie. sphinx-build(or --help) no longer are standard error but standard output, drop OUTPUT_QUIET and add OUTPUT_VARIABLE throws the output to SPHINX_OUTPUT as well- support gcc 5 (i.e. gcc versions without minor version): opencv-gcc5.patch- Update to OpenCV 2.4.11 - can't find NEWS or Changelog merely collecting bug fixes while 3.0 is in the making, 2.4.11 didn't even make it on their web page, it's only on download server - remove opencv-underlinking.patch as obsolete - remove upstream patch bomb_commit_gstreamer-1x-support.patch - commenting out opencv-pkgconfig.patch - possibly it requires a rebase, but the problem it tries to solve is unclear- Add specific buildrequires for libpng15, so that we are building against the system provided libpng.opencv-qt5h01-ch3d 1713817037 4.9.0-150600.1.134.9.0-150600.1.134.9.04.9.0opencv_annotationopencv_interactive-calibrationopencv_model_diagnosticsopencv_versionopencv_visualisationopencvLICENSELICENSE.contribSoftFloat-COPYING.txtflatbuffers-LICENSE.txtittnotify-LICENSE.BSDittnotify-LICENSE.GPLmscr-chi_table_LICENSE.txtopencl-headers-LICENSE.txtprotobuf-LICENSEprotobuf-README.mdvasot-LICENSE.txtopencv4/usr/bin//usr/share/licenses//usr/share/licenses/opencv//usr/share/-fmessage-length=0 -grecord-gcc-switches -O2 -Wall -D_FORTIFY_SOURCE=2 -fstack-protector-strong -funwind-tables -fasynchronous-unwind-tables -fstack-clash-protection -gobs://build.suse.de/SUSE:SLE-15-SP6:GA/standard/934766769d5473d20831c8a73d1ff5d0-opencvcpioxz5x86_64-suse-linuxELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=66f149d751b7135bd08ebc5e5273055b9fe7be21, for GNU/Linux 3.2.0, strippedELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=81be8649baf9b7a3b9e1464963d791c88e04772d, for GNU/Linux 3.2.0, strippedELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=29cd99df44b43ff7faf025cf0ff5bff0302e7604, for GNU/Linux 3.2.0, strippedELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=fffc559e3ff86f3cdcdaa0592509561802e9ba1f, for GNU/Linux 3.2.0, strippedELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=27aaa8c4b9681bb2cf904755e1ec2f89ebff088c, for GNU/Linux 3.2.0, strippeddirectoryASCII textNon-ISO extended-ASCII text, with very long linesASCII text, with very long lines'4C RRRRRRRRRRRRR RRRRR RRRRRRRRRRRRRRR R RR RRRRRRRRRRRR R RRRRRRRRRRRRRRR RRRRRRRRRRRRRRRRR RRRFveeWY<1ԦlЋutf-8d5d7494ae2011c285035bdf3711914057b436ee616d25f52896b4097bf010329?7zXZ !t/] crv9yz_ʵ-K(3{ 'w%豗 JΜR9Wފ֓C=xzSus<( mT9@Y\2kQ^Vzj 7HRbepV:l- 2~d 8KSlrЈhT3Kx4(@%zjȺ#se -wNh->JOLsIuf4IU'Gךs*!nβܛ5V*fd{T5bMҩ%s7{X(L/&a%|9"K<5l1]R)j1M$ dC{ANWԥoWX+_)Dt\ZCԗǹsʳȏHEVqv,2uY I4MLw sۘW0#v+O2KOJ,[_jW$x]^idBA  Nip%莣]GdES)ԁ>ުYvt@]օ}(fq]f4UƷmQO[QV=3lWπ|]4,9GgdtX8Wenhj.CV]Va.%dP,Q+rEJ4\+ABY9)3Ii䴤LoEZ#F')JDuuqA)5!U/j3-9 dEԨ!H)h@y\ lFn鳁&c5zz\(FɮQ"$Ȫ,YG$^$*Mª-;D]L 2+>}#~u ۝+ž`t} C(4](" f<,8[F/kxHI) W]c&sbIwu4L+ q8j IʭQ۶^h`F<skN{Tq"BQG%HH GxI%Ɣ ?0X7IFLhu0[I5~4o~^a4v">ϾyAx'a@r|r(v p:!1z=4rAS^%2~G4k=f❠ up}q9[f]3W},36̎H~G(#_xTU&2FE+Xyj|O6sE-e]->'S9, 2ZMԃ47%IVI B*@od.SqI^'+F(-a5t㳘p9'Frtڌf81Q?*kmS{TAY;fƆv\EOiE,IEbW?/Sڭ&z]sNMX6qݗ")IG|&2vv۸yG⌫. ;Ap/~?,fEyV=N$u 9mVeZ~ 3sB.Vzscytv"o/lѐw%FTUӣN_ݼl~X8ݞ;8/Z! ÌMW( Owխ@03m  4rs} lxƨV=ԇ/"#kOqZ3)xId$RUGC&hڃEgY,ܱ{_Iq[w}T$RΚ`%N@[f^GB_2WVy"WF[(kBmD'873B`̜U~:פ,!_=4,s |haH*)Tox6G5mo郯iNjyEOF_Q6R?:2W4,!P}-ARO'Z i:5nTļNI['7A\zK'W |u<&a i9JOKx E ]E,F8wނ *B^-g$ Ӆ};6icPA9<޼B=**=i X ӺtKƽR>v2v1d3 :F|OA3nA4ut^'1c[ޣ&'4Twv$=JpɮȲmfu-˃}࿙(+P7Z# $릑\i!hɽa׆@=G} eX0nrTV|' rEWE'q !tkUnd0_/;lSKV UU}̰۾dX^ K5A0醠}ycH>9ɻ/|D]bf>.dWdLk?*n\9h76h(A52Nm .1vy#UI'06;1b}f 3w;lW,l4@ZV-zђdƇaҿg;j:]Z9dc3ZȈs]4SsIUz 2:,֌bѳ]vU5ٳeGn;*1DQ ^! ;k ;LFڊÐ/Q=Bh9kwh넁kNe;gpWl7Tn#~}G4v0$ao~c]L`h> L% QY\",8L|g=j)%vJl~VϿۤۘ;e`y'Ю{If([˳"pሻb:TX%ZۡK~X)?'H ʹەo'D5[[$LN=s/ ⬫͂OU9 /R:XR`Y# Y)#T*#=ak:MH'}k`Ct#Ot5`qE?$(4BB\i 2 (\ׂU0ߧ\nAG98J\b be%ʒ1LTvNЮ{ȺyKdU"̷D?#jvD39jwPq ͒vgYA!O"l@ ~h8%D?ݟt>~qyR\7 l:pb8'lYo'PMk1WKpL8*AH':bnOMH.`ڔn\GU&MА.ua%knz43eҙLVF޳=v\gvQG5ȫovOXnUc8sٳB8 t>|ƲƹtC\NnEѲ"6_{ - Ib(dqq\%v#zOǡYU{)|WƳŋ.8`}Lܖ 0c-@ bE@f sva*u5P!Yb35!A|bZʴX|-<%8BۏW[{!ll]0iFcb>rX9|9փ15:(IM4Kߑ< "=#,i7|t`F[;QbMm'ѐ`P Gt(l[K_b/z#uL I$*v1os;;  VەF @| 8`ǘTiy6KڲBY BKG/Kk#:6f9Ц_,|0V 2[ݥS,i9r R}|=/lB SzvYxBSs{ Zrow\%CiL9Tzr#x3[¿lsx]IwQ!|XT巀i"|Y+ų;"pP!ޖ)Ev3`͟gK}yjn;R$"- A O +7V$\]:9>o2i/Kh|R$rayǥ8R˷M}6+> ߍn 8V{SC*,f?Y]CbXꐔt5yi{9~iծ\T %RL-"g!"LW{̅LnvfY30Bd4|)+6b˕#RC)1y)?V!>DE ,"kt Ckz7<$}beEUA`ձK庌;pAc0(O5FӋ>sBͩ%< ) ~Nކ,UC\Ȩ&κڔwg3řa(I7QsTPạK̷1X0d[$ĉ"ꊕ;=]rHsI-& HiKtNN\}}~u5h1^2@$\ *ˣhݼ":ߨZ= r<ߍ\To-'sv}܁,?ٻ_%jnQ0 AӢ 7yW *$1hQ>dRSBE2-I~-#o o&eOJ_/߽/#=>ʇ,tu|g̶" Rў04kon ͤ7+E\jf5SF#ldocao-5b *՞GXMܺժoź%-N9@ݮbm\kA|%(͈nKZRO)|`EtH>Pk6CKa8;!3|Ed$;PgZbTOaEQ["I& n=rE8rB_CdkPyk'[r,z͏<է3Nj.&~r%-" `:26BAy5t]GӒ0əx^Um#Xp)@> Wk`:ljZ ?^.A'%rp&%gɥ;Դñ^ٟdT U竑 -Hmn!k1<>>yx*pL8KPlx KFpȩԏhMPp.X8*dޜ"}QGߜEݻ ݭ ݁:>%Ide:ջY;=9D /)va$|@G=ݢ+>v$[;Gˆ8 ;\ 5EFd(͖!Mq[Cr@C xGm7,mzN3J4tLb?X:\4}R)7P$]aL\4%H"\*t8!MYk<΅;ӕ)f2LL‘g4\k sOBg@Tm*(\_܁Dg\ۨŽ#@ xY(=6˟fA`c` FV\s,¨d ?$aVi(OZ/i[!×-y\౐-I}/?nZ#Y?M@zahG /Qk3[bunB!N2O7u%1j X#˔Jc1M^~my4`@|S3#X W@mgd sĔ;rBTD}+`3>:d]6C{7 >,fko$ [~G%JS;n<+?s֐xj [Yc-fvlTܫl Z3[=DXρ3Dg0J_cj+v :;u|6_~*I׏'WiI@{Bt8 j.b?jGӁ+؎9Z/hi"oxbb$)d(5,-5 v#oC 44M85~Ӫ: FkZ4Pa\ BǨш&g%shYDa=*|"h3|ŘdrGSHsM{"{SoǫLRJQG h1n&*D͝oqC!CT>n-\C=Z+#GRj#O lވbDTxHֲx=zBc9YC9{R)@>*BenGf'JA#= :"k=̑0}ƇXغ8K_~/w^,;WQOiђÇ/_=' A[]`Uj~[amR{.!u-mn=pw.QE4U"Voۦ_[w{"ĵ5(SXj^T{ox61#Yq:x#JnȩdD \] \I! ʰd{wZ7T})JudR(rP: EnOp#lQKcbؖg?6G{螢82N "2Wd< ?PcsDZPO /83DYIryAhwsj-BHאi)2 c ٴ;}j?' a2@w=IfĜ/xIP%nB…d$һ )`mpXZwCR _Hj!רLת3]e8DS Iy@ 3whj5 uu^CߡZ?!CڢB[ܲ vq8.^+,}xc,ru@ZwD{/Oʢx{][ξ'aXȵ5P^&/ЭHVvyLUüxHQWQS(ymڍ& c0֝x&5.[HUu#F ӑ=K.JjѹcYZM r ME GahM{Yl& -Q&2䀑d*GowیB idF1oF#kls`Sm2kl^TWL Yy#5gA騕Vn$4^J6#\iEk4`L#Ԇ\oFOQeDWnt`\ _pEVTij?[0L|, Tq\'@rKߘ2R_0gkOKkpSu|Y%|)^l%LѢd:0rի")?V[cDނu>,M+V$:Q2<]٪oE hHìv1ϔFۀDDi%W.cRiBֲlbi`v$DpF'HAB!5C7Y\ j#] -G?O1g;2->l96.YlSOѽD;e(+9dbo5q"T{%hr&=4r6\{u3-[g| ^9K^+ Vo–NTH'AH,n=P-Ű9߹;f,md>~ ;Ǐ0SFG 2 \WNB)'c#a;39;ҝ̧$ŵvjη92$ ʍ!B1:ߡ^G fuKt|CBeKqQSA~[u7*qWlU9bZ9 Z-hbJ$~&.hgeASp5 >K #aZS2iyN63Jb[òU9G ,"#o"{c[NȦ\{nY!c@fQq =e"!2rHYSB Eb%]{~tַQ^1&p&KD(XlH7)#Cc쮫}hj!xi?aP<@.lhQ{]- +LBvK!Ǝd.E 4Mj,ǫ7"e"㹮d)H|R}}Rvu~ =➖~ \{칥."@j3Dup6P\ !SsNU5gEsLzl@%%Fyx՛. m){""j(wfc< MxtFWWfrg5֦с/ #nĹzaPoYҩP܅`<^c twcsxd>isԺ16WUc B+s"V%Lawv[ˌ֟F0qwt!SthWU8*-@7|{J/R+燞~`#O81Owݡ."3t*:gbjKg432b_ʍa@ ӕU5ܻ5Pb,SPmk>ˡ xt°5l`HJ(9)_kk t/ J0 F s/ox? GCV-(,k\oB=BCyegLcSѳjR'iAjHߝ0;zک)n:'yvB| \xlK07Q9!_B8x-nZL*15wVZsy<glΐ^Xۅ9,>fǁyL݇ TrW*g9Z7q֪=IbRCLq os ) Min lS gΰDdHz Խ>gF7 GI~W%1hi ;-*Ayŧw!z#WWO1 7N^ÚE8H icWs5q>y[Het[z&ı <^?&z6' ŕ%&kF'^wOZpfXr5(^oSJ,WZ~=kӕaqQ`ׅt}Avss%^a1TCE2rk8j `;_(0c\h5m-YOr2 `{ǔRUj֍ObRxl/@-M>G]O.KaTh;JPb D;Ku .3-s Yi'U ERKsb[E-6<ͭQ-@4(I.l#ȝ&ᓛUts/I +LR\FQ/oLGI!nYp2 H BF~܋CNSh.QsKಇoƴk'E.=&~ӆ[6J#B\/$ۄ,:7߅okI3dVo-LՃs&l' zFlc6 t_CH'\:RaPh#YnOq䘹*Gر}'m8*h2 W^{׏~cK29j|ыk/#$[H=v;G_V֥,` ρꪩN d_ܙG$;^UK"pzAH ؕW~#a?J 1$@]*%J+P6%I)}mgbə iY I! $Z-Áj{?"X,hlWYm("X,gR=WV-ؾ̀?b ys?Pͭ4C=;Eް=Y$J+i#t ^":4Rȱ9J%Tgt3xG )*Z=,LJ# vH 8Fl-t𕽱SyEdW_gm..Z7e.ng͚ErO {`|Q.]WJ,-;8^* O. +_OU EdzwjR /wܹNPWJ @xI^,W‡Sb Jpܞ#PƋ @ =Xy˳0 x:{ĂPUV.JQ.\YG4O) a[ R(Cq^/AKhIWdX;ıGLBldGf9nӍCc DÛ <&:,K-lv5O]ۼZ EA,tf` G쭴7މ A+|&Xh3䓟Q+TqH>inT0cc6/Lqg3SAmd=g\ۮp[/4{NPn vG0rߺJ'}b"D\n;Fabv{hCc=YoO"XnG&QmQw}@HOU? ưwWEIA-svAR-~m,ab+~KȽ{O( >ܠ%AT00 oքp4 Pn+v9! 5AcA\]0քGjIjV'araQ:||_hS^ɤww3=4v"Z S4ܽɴmmmo&+Q6}=q^6bO: xj"oS]Vs)2cAvB1O _"wEZ5UZ l>J7qRwuZݿW>USKu窡pmh/$H۲~DAc ,$?r;n&V89дdiS6!f>$Y ~#Ԍ.ϠsG~ldVW-E^'E]GJj¡r0QvdXcvJLU3 #bM\租5Ex_U Ypm'B#>0gқl2) IL<\4*4'vm&,?krb+lM1iJ%$ͳG^LLK۵q3$2c4[.-9ٲX gC3̋c|kq &&%AaIAhu*@auΠūBH@2{!@5U@q렆4{1Hk8;=X L;= H!LMPAy"coഽUզ}Ƌ`/B*CyK5@J+YW`jd׌;&c:ݭ<;֐j\!kǜDU~@W DtMH[9^ǎ(f/3"](@_$Sbv,O*p4Pt5E( ŦMb2Ms@en)v4umb+!%sxΏ#VL})cy GtRaA,r۵4mHs 6&ݦVO3 68N[vg\tdC֯1y3iYi%D8T?_ُw=_J&Y$GU.|w.e]*QtR r>>IW@ Xփ K+"ogY;uC(fڇ![w¦=+n(9ѤUBrسuI0ُvA5͸]XcW _9@:`Gvw][iݼȘ\3ō֠[cTE%oD$U*b@ Z?Ad`.TK)Vp{w!d㒄SbكHRtsg]Ʊ}~]ËDfzl6;uf\[՟4i э6qc!^^"z^9b7$&߻=ՔK͍ k90|LT 9btE*Dq ݢ;l]V 3xXK!]C vq򞩾TBx-AHu[s+(ǭVhتDĽ⾔Ň#$şe}sb~i<&ST\XN;[;p DE~H!c0ONg9"HŸ#pMy2AU| *g j$*^][N%]y+5ۍ'lM& wqG{&AFHKf7/2gX6_G<:[/ dEOZI}0"?y$$ T%{ΡfxoJ _-0Ѿ}oγHS,0J?k\lϿ6#Li!q\LjKKx@-8c1A&xǗLrO5+ˈD]VAjB=[WY*^ɳ t11vU4a:3IDP #S qcg]&x؛y%Z&' ua*}s4j 4@!C#I )O4$ /ogF$/d2|-(δ[$-"VA'I.[:}UX-.qxi~ >M[1+`{OzZr6{j/9l_ޑ^Cy)o ~NU ڒ9΃e9L㧙>W)d h`x+i>z[9 yx(σGvo]Gٔ!<"[M±qʹ≹RKFP0A{u*b@Ac 5  <mM Xg{;JefqZ#:<,G]R:+}.r$O? EGjjiIYT̥!W~[Adgt`rff-<WxsiI-hk@_ \U;7_qa~vbE k0d틛^2&i-) 3Cp!e=տenp 2m0 ,%TD,˜tˑ :6N)tvm4=  s@v|B6NIB%0p22vNmnX^I#iK%?o#Ln8xC;6* ?E ({r4FQJ_< ovXLC5E|iI(;u3 4s]74'8CFpg[<>V/Sr$ɾ|ҩ*R (&"[ ;M ^LUqyU> VU~w^Pr$EG%oV*Q4ƺ[7t)C'^*X(rc9&3$Vb{+}GOT QI-J?Xn;/z[UݓtSp 5j raa)嵁֤YDZ6,ÆQJ啍: DT81It {l+oK?[ۏVx/>{talWg$ wUHJ_3aL  -5KH=Ey 3mGѴQqpD sZ` q:)Z)]tjù/,8EΓ Zp^7dzh``(ѫD}M[*jt_4u՚9F:~~`ݖр,TB%95r"QqşX䩓TiL>j;9 &U"| {ϴgYJTQ<٧J PBLkRŲ %Ϻ_\C@_z ̨VPb;=;-ڗH]">#ƈRN֯먑g|.ɗ,tqʯ\X ^פ=sw^T:3<ޯq8MB1ϡ ]W/.l@kJi#Oe%gф2F0ݘDWt 6Uuπ]?q`], "PM9hdc# ~ǗZp鰞M`j,`;g]_퉐UY}g$Mge}RW8G˿¨$=sm} 8euaƲ#s֘EsbЂDx"v7=k-k)Zeuޔe'[ JrcQ'$1 n#17϶p,'[*ml%Q8:O./>"困R>=Ϧf|.l!Lrd̆<0kcQ̆kpo\]kam54.'l%.dDfPBY*m'LŽ,JsKDzdݬwz+M▿A^6WI{""rXf9.#% T!?O2~_ S0\\xYu~ZJ"Nx_~/7not ^rp RhiFaӝDCE$]=854()qf-AmWG* 蒨+N#cRUeЏ]T%F9%9VOGά kwE( [^Kյ2Jǐ:+k&^pLm#g'`^* <ſ\Xri1HEC"&C5,iJxTpSUa,wTٞ)+L疰:>VX̐r`N7< f2)~1O>*iΪӶ bL:]YfB< cqjɄ1  )_X|[O[ I:~%Y׾Ј"rwĮ'Jbçc<3 fk) Ԇ7z > 3=:{h%?/]H*&]W5VI~ #%S+#eKLnL*\=q+| J#1;p̏2 R_t x߲N]22lDZ1ʑ֧tz9_TcBIf0UoU'ԬѥNo f ιGc^bF .&[]aL(с lJtJw 㨩,AqxlHq êKUlB]|rGM &ov #Ls(طTa2;T8܏XInJ] W0=ubM&\:K0sd:-4 U?q9[%zjyɆ\ ݄vg~*}{e t>-nPd#CjeػG4 )^ ox f*=xP%1ïi_}e>UofUZOn>5%W^ZAzQ9yjJLCRB#XMB`qS_n{s_?Q.s̑Ej`!F-LFI)@Űd){ߊf $9wT92=zFLY@xfhXOTA8 xT?otO0TkP_`5=-Tql},t&=?:¿;s-13j)*s; D$HNh/D,D=c阠?,iu8.8%B9Ոb_J,$cɂxPFBH#jbёN H[.'m1a fcMdw@u: yNbxLF w)~ ?c1 !gᎃ+q^='CWk>^lVu\@EuЃQgU-[Dۋ wJjsDFW4o8/z)Fj981ދ%u@!2Z#tQF2D@h fWGț X$-S_/&#=u@>8#[zG)fks-V{%Lza,0 : z)$b9߼_]T5挪Bc -#6huPl>o5?)g| S6~0sWtдXǼfFJ&;NvCCDǻ(@^RYh Dz~8k0jz*c@)4$d7B%`/@7'_MY-X 00Lro4#h}:=b0܏!~4Z``/f(+0P{gf># 2";|co]e21*LYw)A`VڸKddذ@G]9!˳sdȚ-b䤰ê0%m';^wXw>_wÈ}A@IAT'S_ \l_j7xGY̶ pM]'Ei:\E@4 +ҒLQ;=DD~7ۨδĵXrŒ仡 X৅7_UJȵ$||;3wDD+*wHj_GbDaPW]7θ=\lTB|p$S~(6ըC'.J&>:{24\|+5ҹ5Yg]%uܷ4w䝻nG5Oم,e(r4LD7S3lo Ӧlw=8)ϊgx~A+c\O9wQF@\ ڦbO z&ƀ'QYge?^_C YIՅ.Fi$ ίhVu*͚9^g6* Pk#^UM{OA# (vY$ԋ #I.FXoM'uwOP+3|R[YGc\[Y8 /!'>&J0#D7"΋߆RZzC$QR \0dZ8>qGy jgۍp)?2׵w/ o~hAEٜ^Q%c,^θz̯.#*NA G|Kh3u)& TD [7lf,T4G'ꯊg^QTh__ar g7f|/ )P2 R(2* Z7 \z`;R{"(4!  dk#=`Cdz UK=ⶵI]4ذaĮ6>2~}LDm}+ Ɓ"Ri)l_.V? a '!Йw׼Ϩ<^z2R)bKwq 8WJͧ92o;"tݣJ)9Fq$NØĭ3PufmmRqvko [Ve>"0jGŚĸM:eXal^rN]:64g(@:3ROP2Pb" c=j5U1%UPM IYGic:9 ߶XmoM Йy=ہ8.[x!%,%x/6X6̎%9SFWL-;e m5H 'A@h'u8S)ͬ`)WLa2|o-xY46.Ng Ѧ 6QY jemxS'5FكkmYyEvկ1@ZhЖxC+5h!8Gqg ߚALޚ8nt"^gݎ0HoO/>w XAP8E)Oڽ3Z] 4E;-tHQlޡB.U^P>oCZ" r<M0qL <^B 1۪FObҎeH s#y:Ii}y Ae4d\!aS/5P+E{ًgȌr1Li5!yjh'z]C,}P:Z4=e6}IRlӵiw̍4l#NX*&lyo Zڈ 9o b%].(YN](3*f]8sy]+gk*wGpI^龑Y. 3z,H)*r.Ld}BceX3J~#N/>=țŅ8C˺ :8swJi JJhS{@^"Ȳ jibr6K\.Bg}ѓyjlŸݖq\x_W 2J ʺ_` \ڒzgzIFgv +PQzU" BiQ u.(|3rxbVHru]t(DŽrnnmsЄ~ZynU2Q&lXwޠzukFs@幈4-x;IJcʎ/-%/]'ǵ8&솠vSy+&|Mad,*r4itTdrXǽŁnxZ %e/-wRkOsz#r/ #bd_6`1 ܨb%#VPj=LV.)0vI a1 L4JѱY㞍!+aOu9z`\oyzEpC6P9{E StP cyc0`N=J _cFǫw.[(viuIƁ#.r}?nkð.Dx=lp>Dm IF b,>=aٶڻwfm1! Ln#1c{IݒKw'Jb.u0BP _Le#2Qsy1tlr9S00H{Gfԓ%fK^ދmךt]Aϋ/b)eܶ& nῥWW`yVY}Ft7l, Uq"/AdSV}mzs_ /tpG&U(xGP=s3n@xof4=]GV%rζM[ffHg&yoa9LOnU&]MGA糢"T3im_1ꠋtjr |S(PdC)W4*"l,ZMuJU>u?E='g' _N*6ҟW饈;STZ-qO9g@坢A5УDrD M9Xy9a䘷8PR͸4%"g~| S1ށG,FKM<\9_'\y2:ґ1)NYжu:dž\4=]=Hļ3qD>_jJ]-+_Tuy*ZWh,8&hMgR_d?][<؅xh )HwҲƘ5DmM|i%[aayz( mZhYΨp&5!oEnp H}%K"}K0z49^d';[$E  ϪZ@^g}||cEgHA*^':0łcSR'&1Js ZtW-G7fޥ|~u"]#O֎~tV$i$pW|R挴t1ɽ -5<ϓ$oQ .jk3rWi*Q8@@5Pz rhUӱuGpv3* =G؄N",B( /7]}I0J;e< ;5:k?ҮU $u274^^{'1aN{/;n2߻>J=ٸٖ߮aG7U2ķٜIGHHsqO@lE."qE{R=+/գ7`kU {YrC@J6Slx+fYMj`;gս4VX#6ϐ܆4Nx¤Fm{o&QnEnL|Lo\$Cdl=Y_|Qu1,/ƗQ[bF!H~_e"6`z>}=C g\IksQI"BVl"_zY;$O[+ XFsH0>P V?]E1.a{SYc55 4xvA.M)Q%7A|a}"IYy'80GnxnJPC{4}V}v4~{} 8PfyV?AKld|jɯ[h,sɚ\hj"S(}53( E{R 0$km͡GK|ҕrOyut]9f8n@]?*,IS Kjc(FTsr솜$K7 7l jݿE6qOzԡyi6U`>$~nM<`w+MTdϊ;+_eREyFbsk[M QgޣFͩ:53UQ{Z8Ow?Þ7=V,] duT5#]kwNFH3UN4Ky /p+*K$Wt_%c_V&!LU|eZw+[{sw2W -sTDh?(Yb\_&Gy-`@,iOEp˸`gGS_ T#xT-DbIObc!H$b2& \sh{Kֻj@+Oi8uqL)&)W3횼aX?%ѩr&\cT֠q* H1VAGw zFӱt9W\dP`Ş[YCh|_jW^HlJU,K'N6u_q7sñ颥&Wùl:N;.Gp^_JK~ Fd8du/7cjӀjF7_+=춙~wS3VDvB AR҄9-Uhul#5 Nǒ#d&bGǬ/>ë7+؅Iumʬy/wTDG+fKx|}# ۢo!AƃXCs{Y!`P OAx9rϺQr2a ^թ}ROq-0q'pRE`wO)x4غ {8 CcD~" 2|4. 栢M\7{!.rʓD;2iOM#ߑ ?!۷ o'\V0}ex_oӇFSO$~t_Gp'X6\GY(RVб#:NPzgicQ L"˻!F'8l#Q`>sAܡ_i/ޞxJąI WTMn^]QQ9Z)&CfJ"lwoHb3P g0t qǪ7MU+af)ՄЗj9[)Q$#>0~oD6 zڨ{*2a%?5./h`sy $e[ObeC4/gq>TWshأTURךvDMՃNkaXjLqlyBWx`Fx\2.d/ Ēn%z~f O|l&XnSLh x'>c"Z5%Ϊ5,l@b=w)$ k s F$xw Hr푍77n,x~l)xH'Fx[ZI +q؄o: [&[]Q\@ojAV;U3VOB棠A 6eu%䛝%2o:S< CT[L2rzQu[,HE%Ļidf[~%yg5ϜB}4˃EN8 G'f"ǩ/ǚ$4:tPexbRZ0θO9_7W+aĬ\߫E '22!9Qh뾍1Q S ⥧U\JQ#D]CS!yP7lxW_ÒvOC+<)1ņ Vʜz?.D?Г>c䫪Jajq.0^'%sAyUT7.0IafS`|a|Rv}hHOBȳӔ1BzlνgľgJȷin!w*T!jfjAX;G J"QJ:L>W ǖ*"?J QPn2~iQ0;@u=^3&ZKuJf։ |9| Uu,\F  *z=5)1 QB@C(L;78E<_~vL[ Pfc E-D4 _L9"g 1? q~[ @^Bs1m^).yOIPvgt:<8Q$J#6KYCo3u$xS d05)PT=R$FK#`Afd,JQz丢6Mt~pSu'aJd1<,N^dP5n>h>v+%0JԉLfʔj)+ENPF|-&KeOiXV`:;^Z{ O*vP^x''ىlڄpBZ99y.ߚ:ap%EVϓ4by1@DN0w?tM{68tY1UvWՠG)'Ll%;[ԩjxvr^kNj~>ݗpYϷ[+b$+XK('&״ʀSI̟ r%%ՃB+9<,BŻs?ɝ$w2PMi`FެspwזwJ}0)26>¿Ӑנw]̿mijjGSlq|nBơ}l#ϭor$tU3/=icGCE P1KU8=1ڲ XoBYm.A@bt*i˻F>: KDŽR!RkfzG;(?Y½vq/C۔hMEdB_y:,v{Gfyg3gh ||Fߟ|f5)ݟ(vdc8^\3Ўu"ޕEbZBe`dYӏ: PObGSߛd)*Su![9Zg.q&EXk'Df8nV$*x 7[a^q$^!q"e'" 0bw@p܆LJ™Pڽzi-=Ak7xϴh ; ;3Ę2;wd1e! @XōN̷FaPK%4PouM pM"4ڤ sԞI"JD/93 ߅-=FǓhni +:հ#(?Ț`&Y0qnA,~pLHZк&7j5krЃ< %eރ8l]hKH˭@̀gO^ٓFۢEu)|1.3uLߧ.9hҤ7H=`aDwt隉$/.;HmY I /FFpnp~]m.}6 ɴ>_s6e󱆈?l#S<y0-7bmlܨ{xmzOKST,6WHN (} @U,cťv{WF 6W (zPTHȊcr-0\7cxՄ$R}|pGwT e꽜OZ{sτO`֐@3ְ%dW&tce8R;T "RJ o;qG{þ.n@H_ i]oJHqm㽈XJe5ĕX=1f d(Wd?x݄yĄv5}䎗\H1&D"Sb&2i3j^7Ӧ?8wI7coﴮxPRmw>"Dij\H=bʐC;.ם6[qZs1-_Uc_ =fp>淜Au0_Zd+ھz^&u'LܾՄ>'N䘧XRzoa{0B=D MDC*BZTZOP ) rO ßm4~2tw5CzCDR0ӻwlؙ& i:舄,Y0̕(Vk|-dm.cE M,L>WeQS n~hѓUR<\e 5_u soĜŤT(*9TZu1ړe93{( czݯsMB,̲nNT*O|in_(U #DZNx]ٓ"n+(']B^jKc[R}victY [!xIbЀaW5W7OA0S`|32U3"IdAQ9BmWàОp#h;i7JD0,5c\"_ ËsOF!/_qkfoTt~mmThE_ d?J/;SHuE2 })+БH]I>$$;D_YיgS.kMjᣚ8~-5>:*[#hanhtɠ=zT+7OpP&Z7G/Ahj f? -e~%9,u㮚.7B>Lh4O7"#ȣ"=dgT3A&41ɽG1މ2 +;OFs[N]VufGy)3:zg9Fu]RVMg| v`eA&T{ӱIVk] 9"xDϤLNۻVʇ\qJ{ X /1"'rY^ᰵ s }iH*g9R s k|H 6 K.-^uF7})(De#ԑxzg;-}:}`dZia%3 *KGY*dd"Pd0hhZ)$[ƣWLԴkKC%(O 5lA cJ8]bP@8z!m'Ƨ&D2.ʶ})>:J8^ VaKM+=#8/~鸉kX6yw fA$Rs9m&ԑt o˗â"bh䟛W{݇ӍBe~^Ά.Cu So~9/MV"+Ւ"2^H5 4tK~7n),3_F"8;UD )"TLX_[NvHkۄ/F$7x٫3UO,砭.ݖ/1K` [/KFi,_.i\:Y$A*֘HubLQaGfӐQ'ufۜ}Si'mHJ^u(=/~ CZnLW:U>G<4# +HwҵTxm< =s*3F|7(b "PdT+K7Oo,zl@BIV,wU e? :& 65GmPp6]q*v3 5l2rT2v4z>txWHsfY ۩vt5| ph*N[nPU.:1xHz~;@A%+}bwVlО% ,{ڣ_5erXRgXoYxge7FzWMj1'>'0gd!gU ѻ oC .&;;}alPL֨#~mC2X&]V8_1S1)T8Kni!Y'š}f0kt— Ngpϗ q)X m㇣͑Sz}Iq [#QFk P70δdZcr Qsupa;dfT`83ľt D†nѶJQxߩkaH#qą Ui瑪91=wk8V H<]l=y}&ò};hf|"w#HZTbRUؖW@k .2} )wŵټ۴V ϬN#\jF_5ZM-R`G2+-Q"j#J},-ަK>l%뗿!1~jjj(jyO-z)"֡ag ccUjtl1/Ƕ˖wORxlY'#[lQ,{kY\nI4@ԦQr*CYUfFւD{R.zOL‘pV Ox7 x+xJj†TDVDB&([(| [ZޔDOk/lO'l E35吳OIb7{y3``UOeeq^ kC}Z PJU(8t1"|!PDHWrH1$e9cxSR kٮk_T6ⅽ$ȷԓ%ٖ':{ \zW.8nB( rpܷtdGc4W΅>T,4y-^5TCAI ]UƖnkKڠZ,n\ 735ÄN sFblVmL:Υ<Ҫ(g eDžCKo,}{(<4EX*?} 6 7,,F7@Ol`@<8QmoGÂPm 6'X?m\ ƶJN˗Vy);^/{ z~yz= vȧ4H/?6M0bå~\fWTGQkA{uƛ a9q/wӳ'oBH96*0Epj7ԗY T ]/D/RRɍ./7{!(9a9kF&a4a}]V1g)l 4pٳ5A%""rIL e I/YڴG5R3=&~fkR2]PmrZ%řVf`:]P×\ׅD%'fd+iP~F J`ʔkMG n&&E=n#X !BG_C@sY|uf6; ڮ-+tlV3èbS`,\X¸) *39e0]Չ@ʫG RźxhC=b'\Uo7$2DHXrpn>Ćn`8JNrvj!q1 dk !oYaR7Ad3 (.\}b?uS216Z\|_sۏ}`^g.1j6Z'`(i)IQEOidQXTl o*evFw `aMh*n>(q#YCe9k+2h #L_j:#W,KF͔A`s&6@Q=݄`9 Gk`zTH:%TM[%}#`*sV1Z-qHҢVR&sTdbm {տ%fQJf#3ɢ;>j'Z.oO}_{ !4ͱrFٟ~0iuņ?@`kF|BVcԭo\@L,$. 7`1͞'bH5ǵ6 jf&so ζ T?fV!ZwFK؀{1O"فޘr>z(-Wuk5raQޝ@=ډ Ekl5FUY5^XߧdOK# E=8NWq@L ZHJM;19SHC!1=֕ mNw pw1%+o{U)T2./&h_hE,2?!qƻ)ANjR xC#\'wftxU]^rVgIBMvJx1ܻ{Z7k?5ٕe cf*:Våȁ0.!qg۔ֻ`$o[8o7K/Jl{+]"a2vw)Hw(=l񜈫FXt&Jzavw&1+K$_/y{))g'G\lXmgɶl4ʢo appNkՊXPO;-{GsFz1ЉPًESrZ,fm`.MƟ;d9IT4B&%5iz0HQw(Ri >ZKatm%kޕ= қz._ѧ) ?V/aLX<>+ \8d-RL0Gvh B( #'p` ,g}uQP+3syer'w,?Mr]u<] Z=q(GR Hx 5EsR 3A30@$ P*$]4 3S)jNkVn>BDK$4#>Jx'n)'Uytk%*s̺&0)JP7(UX¿ռ$. Jc'$QCbW#ջ\,;$Re\ǁt 7}oޗGtHB5&Ӱ?W'sE-5o;WpOaT>_O*B&60>YHs͉4 G:d;ŋ-@> K!4rzU:mݠ5sC,.kr&`ADSgfGAtOPW+X4$EdU+^ӍA;NtSw|P<PA(l@W,40rFVh2ϸekʽ 'F-wh}"9$B6) B.n 5' T.0(ak((| :6gzԓ+3gG˄f<2((d"#g3;ĉ*Kɕ ;8S,Ɉ5c1ؤ,KU8kP )ۅ⣝uNՀ@,&@k7; gywfH@kg{ klsxـAs8Y w#pLVrqN|g=tItD_I{W3Э>Z!!NWa.(#ꎩZ )iȞXڻ*V(%˕vӵ;`xOޜO+mGݬa<ܒVS uS)Z x#:K:%R$_ ܭ 'HzYs&g4,,}iAT\=_,gf!ct^bPX RLi0Q-JtOy~L~|V0lX5a1xxnObmeqwϙ`mHRʳ6ْ(KHuֹ" +yy4U/(qq&ͱ!\AnBvJAM?$žXmeK_>fV/Id2D\ գCg&#eׇQ~꠱'Wr3ٻ2N,, dG_}`p" kssNQս4-JB(AZyq ;N3MzG15WfLY[yr2r1^-5/mTJGW e0>)S|[3!,XDa ȁv 2Bh\Iܳ.ͭda5_}4u?*= >ž܁6dŊJL+Y#VϷJޞ>,@=ST~Qr-a܋CQ(LZ׺Db}lH˾8,fHvD{+A}E_D8޲-U,%-kR.^K@AcZюmŞAX'挚5ƙs."Ѭ}쫥S|`Ҫ0JU?y04F7S]kOٺ[kkS#/D:%dܽ$R{(Uի铐ZKOˢnTF2Wuqh)Wdm!ZhF"`=Y4}'kP{_Zs7.U"jnRH/s}BTk֫!eNBw`o&Ypv_a#~PC1_(A'# |pCܐK f${5uo~ŒAlnhm ۧ{SK^C-t@X{+࿐~Pя*~˩g!:Yd v ~OPR LXޚhCm r{JN dak=80VznlC Vԑ w hWw?F!v! /"okα,K}- bH+hڷR֒]ˠT`Zu8J(MfQފGkο~IvLqu̦}kTwm̂+>W5G-ϥݼaUCqi?Eb 'I+KJ#L]E)U!uY(iq vRǷm^`?bZ17{E1:\ڄ ͕n -iUr@B^0 Y£Kgdp(P.q[?/}bx T9\#^xB_b/iSZ"* Uy/ڤi#h1_>C$b:GE4$h"kPO:v!̝i4jه) X[=AJt L]hiN~>dY"RTKf&CB"u! OX6[ETg#"JFh "&0PoY$j̟]WܡIj{:ԃؑaX 峽}c Z'Z׉S9"ռjJT =vέ1b54:=G/*U.@߆`>"&=SEr m~Z .; = :$:VvM5e'b_];:F|uc .jG 4j~,atlStJn=7BT:p<{:ߣI7jqQ2C D|qUԚ1mY[״I@o~LczH,^m}ɿ$Xsa,sXgM}{d!sc87btNMY. ߁ hulf {Xk̚'M>w+bu <,f=-oxPYNmvzR/ L4p͝x4 A1,5S;*M'Jwh(Nz}YhL1'&J _=B+)OȂT-xa!y?;dFuN hF\ QP" cY)LJSFP;LyXud6U>%+KNB-Lz{U:ώs4KLU@^W1KLY/ 4GQSHt+ON{YϞzqe1lLmvnæ= T<#_8LO pWq'ݾ\Lpas %dz-}%FQk]v9 ϜDosB5ơCg'[A|vW* G Ẃc}MyįE!s~-#GpoOʠt\g|v+BD=[}fBjn>oP_ՏIw? /.&K|6%@ N\ UKCy<s-ф׆H3 /p5: բMHi)z%Iᚶ?`Je+ח"ÝYQ>InzNlDhY ĺgƔ)V3c[&DopNg/XWA`ny9SgEQM5Si)lbԌ*C3D2PAVIίEzJ\y&?9nt"1<٨H"ػdz >/ 'F|jO(l)(`;WSX@*f ֨찝T|77eX4(j^f3I$ FoC5J1\†uBiYWq4ݥQ+YN)Ṙh Jy67WCݶY`("Cu5g¬q\Wy9(GV֫v9^צ}A]7ъ˅e|dBä(N0- e7YmE^i^pSlIBF8 -Z% y|#yOڍ-gG^FI"lcsAgaHŭ]ʅui 1Tz-~Őٚ9e ۠f]m7'Ϻ+'8dmڧd -ܚ@d,FPiv~7YK%8DgeUf'^PoPizW^pR\.wUH\[ 5% qΆBC_C C'v0[t ˣɅ+/Ḃ} 5!Z/8[ .󸹘eی\ =졛&ȠKsY#2iG6ޖ tXx׃%EN? pZ6j?X0DOIswd0rɅ,:m;ޙG慝w=l77W_tUNWYx _#P߆麆'H~ݥ׋!÷%IAy26@z{Ya, n\||@9-?65 ?%`H7M R .*u2B~Zؤw[̑熳bXd*^1(Y2"!`"O Og2,w~Ṉkv3kL [,.x; %洩V9(b"ZBe(RgoH:Ok2 !-_Uu:Sb$x6YLc{w;6pD^GUHΑM!fⳌa8hdbe+ ~,ƛާ V ]#ZNl@Sr>n4"!6G]<*g;nl$Ix?F?G:% efObzof}ZBёE|\-X=0Te/p%m/,?V1gj6T^p^|gaT~rrS W_>[^kQ~?H&!\,\[Dgwn/|. EB kC<ʂEL *5)|b`\+f0 rQ!*L'/r[rs-x'^k+K1Jz8o2yQsz&3G|0,X<']>hH~Z /ԪЋZ@52s% Yv`cҒ-aAt M󥟼f7GxL٪ĺ?OeKol B>k^5Llf'O:n9fA" yd<0XzR};qL[a]硤 8Ҏ䏁 BPg" )q;>ڥS:SOch[WT2jp\4S-5lCs2[DjJu2!7:݋o"nKo1M6%}=u8o7 q3F ʁ+\¶iPl1s"5Z֌z_51ճrʓ( ژAC^ EE)N\+nbߨƧ 9~n`r:E^E sSOтįCrΛFW͡aE}W~Q@?%Ƞ]jBxF2D`UZ/NG2DY s]淅mlGyn_,m@x@*y<(u-߇ 'OGEޤ2GA xDh@fn[끕7UG|`3(c *tz;V~#@/;;v+比pR8/ѽƧ4GXEoAl8٭ESG?ڴKP?pmL@vb7׍eǣ.,PΞ'QZP TRL_Q%׺nhWa0qT_c1N- *.6m4]a]Ex kXwPrk Vywa&N:iawx^?`ˀ!RH>Ϟ2_DϏm&ѴizӣJz,Oµ\HCs#{}ʀQvm'_-(M1v`--Vpxϡw]S]cAZPWz[=i_*ʂ0 B +9qui xݠd.ޟmmBL.ķ^1/+1hH\<,q9UIڻt7DR/Skm6,Sֽ.C# p6odoPfʶK͗Znfa3uiH]lG{+!}؈RPA"w=Zщ0kʁ8+%N,vAQ\fP~Ʒ0<s5AbC9ǰ}Umz@>Q )<-y~ޒk 0s,Ȫlr9fͽ8^dF4PGp7-%X] >]DŽ"C~a+s¦|߉H  jO.Yz پ\6`JxO]`$t+&O4櫉jل\"vDZ-2z5hY|H)+,(Q%7j/a)uAȓ!vx"HBrcOf?vw\jĹ n}@9먢_1wcl6v2p8i2lD[]i}ufO5$ U.7;\Fً@CD|gU]exqV͉;O|AC D[XkҞΆ&$cοA+ͱԴx,gmɲ,8P'A㴀[h@Si8`P f$rQgj!>HL&.kVP]$Ԁǭ9* ` hM )*Gs)! wom Hj;`s`SC&0k϶&]WX5t\B%xr)w?DCؕqC%3^*ͨT-=&WyF "vח|p cl<㢗16}dgiE`zp$~Nx-'y6QvaJof/:uIi_cQTԕmzJМXZUfX}_>8ˠU4bOmԶ4#k_*ol3xtcὀ[!iO[8c"uuoElyE2K~aŬݗmS06#$c g0ׇ)r*]oJmˣȀM:T!|G4߄F /y^qFrey-QL+`U8Q^KY(lqJN퍈KGm~v᠉< f%̍ jѓp5B.ڗ^ e965%}rEh 39֝K㬌Ԥ·l3Hbh?SA3p:#~V֥'m3N@aX1)߼h=_ -kBjz|4DXq9͵jA5EM+(wtIzc&vB%r2RCdBA"'NHZj_{(~% ,z/oduz;8 *+}7TZx|n1.ˠ(hqi*˜O#בQ'#[Fظ aȗms$f`{nB{ iS[lbEi~c SݶRhUX>%zXp|P⣑rlUr L|%2Y1gcBo]ՎzIGTs=L嘾YEmX.ʈXokdzNO=~6]90;E=ϒh,~qu;X#v ?9$;j(U5rIXdR~x:M2@ &A'.q"~V 4QA^LOݚc$,w>9*4FP\a7EX:iLbvHrETZux`61M9}~dsDt8{^XsC{im<)X/` Mj`Ҋ">>P< dj&-( O%'YlXc yՔ 9LzU-: Xx)LÿzO8:V}}z>?ː-l)w+hx P}u/šRT"k8Z):f]W@|Ue ]BITV/\Z^nP$:Z4ҚҳW?5BOXI35GOslt!+TOJczgyOGXeЭWΈ{D :Z;7A 2M8 ^(%tޠ4ϦF~3gn{1/L|)]tHx~5l䬯e+9<.Er^JXugܨj׺l6` &>r%Ycasxl#lzR~m=o Bsrz8OK4ս)ȇ"4%T?ye-?MYc޳5?1HnIRx+ Reӻ$K`1TԀb75I9_{ dd= i}ĭ]^\+f=t_&]%sF'5@[$@jmǽ$;U%G:x{b+-Qb[1K]=УcRziwlc\E`Կg~9*9`}h6u "ON?rLxْd&7<{.Z-zQmߢ fxblL0;` 71]A%#kfra/1 oso|nWnZV&io]F%[lAS4ۨE6}nL7SXy+BKo*dCvP jR1dK9j:qK}~ rC/M+,^ߔgªe";Ta)}}a,+i/x?gSPM_Mٻ9xxҔׅ9OOeε7* kD%=bƭ0&T9<@ ]ւM|`%[ KW `=5#e Q[";u ҶfV׃kec RͰ1B*؝HCvzinѧP_¿ݴuk9孊E%#հuVko2-0 !!X\-t+sQqwV,ԽgȷtȐ*i!=n JjR*a3Pp ѻf^zŽ7;]! d/mNB>ejieq3dg?~el}֌e3vEkqʘErlw+Tjq;4_ʅDrCdzӐ"~_gzcH4;m1uGPTikqN$`}Kj1OyR=J?*շ.וp } t)9˛}c3Hf572/bMJ #*ݭs= v,,I@編::Un$mzO7~3" r׏HMKQG@3?@4"~L8Yi b eh7(%~QV9 Q&߃dp:(}+;͓;ѕv(sdqu[T ?d-Ԣɨen㑗wEХrFg&\*EK0NYH/֘R|m H5<9Bo_48rɡQQa?Ci P_R}03(OWQJ|;ZL,J[5w zOeY) 8ǀvg14esI":Kr)-&Cڳ6EpU`OIC,pbykgQ 9;#(jU EgcK:bsɛpJjQg5q V'EjsORݤwRN?RΤa蛟YDG3&@eg >S b,=™Xba FL pjqaѻXQ= +կD "d>|ra I1Jty" qxCsU.qmN䟒\+q =nU)D7*ף|C:$s7Km_K -d,TK i4vmhR|wHwv㼈Dl&l6ՎFti[aMB_wr_S)!!ڕayഌbY# wsGzϹϫqڊ?$TʬD;6j$` Ô|UL%؎jt,GRwJ֭~=yQ7> rˍUnRr %C1&'7,7Hx  vi=t]E6DnYSOC]G?B[4/eYq³mME $W%xL}klZP9'ZV5+khKo7rۡGmV~^q<%9MJn fRzI2[!VA@"+x @#9wa6duد⁦AR7'~j3$_8Ff浪׹O:ozuq6ѕSѮŮq+\Zs m26wh 0~5UEiF!hQ3c/o <7#z/+|#FǞ?)/?"`܏9<Hj[jYo ;3z!377dkK]Pֳ\ݺ^K==*-&|d >H%[o}J+]w՞C!$xCoς+g51W_ i ٘b[MH ]Mp+L7PDm,"%XލsLMe1o|aa46]4O)@'h@TL񔜥ޛ| ewy|~>RVqƵG%,-z4-\WJ =u2އR~4agv!rauBp가B٥Ca!\37T- -;E@7',dkZu؃u"͘(`9 –< cMыn&<3`nE_q\e.3 \6SLYUB{8*E58 !_m߂GrTwIV9p&w=p:̄qStbutc?V qt k62._К8=I`{m>'V㖥7H)}߻;X3"`i<ׇ|'-p KV;!̼1v,I7fJ*r#| v&n>-ht&4NHCnӗMIZ <1 m mA.ny> v7b;M+F^M͌:QۘbdGd)Wҧn[]) yߏb"Ř`gģ(ܵ{t惇Rn-9Xum8Y(m)?M@I2\(e `5$[贞.]qGFG*8pr&n gɬ ev| 5rVң6NX4fgGv~{b,m/#E"DI{WkǧoyǺ0a%AMlGhc=}r^2u(PN9l(vt6.>e!04QF CǾzWG,)oJ[HSwFK=$1^NNeEUŭ\7_3-Q`=='hWz:%wǿ|_R 2 9~{%/^=^ѵ;( jjxiPUΰ 6z喈W ".Oj{oDb ~qkuBivXt…u jQ.9g pxך :2MMv77A װeQ_d("|b^kr3WڗeKvgポa",FȃG$d`C̊fcxX beXCV%}?םǘ 6^-u!vNNHMQ%Ύ%`;^W\fFSC'$\ķ`鯉oT0'^nrW6G-3AD\y}o,rgU%g x|eBC<';6gܻb;/tAhʺJP-H9;Ak^5NF@(zqU72y%+Cdx IlGz,ܯ_pJf ˾gf-&V7 AUIbHJ)Iov;tas!i 6m,_BQRP?Q|wDD&UbWUC5A)G{k=?9[/LLyE8b]'_5ACJ"F2)4ˇIgC{5<'e̤ͲA*%&ah_$,USE#[.ı!4sv]lhڶlgyE.-A% ɢJt?ɚ?媚9}p].oGAL+gYW28YPdPۧ*&F/'#=Oe'x_4#2L 5/j臯 h2>`%Vj]*pE%x}<9 6 '?- Jyj=TԋђJYX_<]{,_4D6\)p#`Ce6Dtz;pq xm~` "o!#]wd,[Cn~q|Yo}nϬh;@~aٳ GZոTFlj+nCKڥG.൏=('rwi*kP#ğTxC.1iW;WX bB9-%Q.!v4M eiHqbV|Ȱ03ry6/̖sBAbh @6zAH@}yl݉}EfhCO+(v$kh;{zr@=im5-F=[ܼ2nθ+" FhXQ!kI|+H/ 0I&ܬ4sEb@tPL{h ;D=Oe/ fSzQ j@z9B$5]jӿ^aWTX^Co@t0WFkD[ #@Qe֢%c2>z+XkUj^M㪘Dpq' ޑ&]1mQp&wS\2 pC}ڨXJIah~E@&Y7` =m kxIqzἸd0{x&@Osx6I! $"u#kxs<'q~;fZa\ b;q)DD#1zc yq'dv7j;e oVKR)D|މcB}~ӽ'68`CY28:SWտ(ܰGEHZuv :ۙ~92 2Jɛ:_?PWZ`da ¬Rz6 -?[D.#q^߉5YSbDEaaB-CLOǂܸ2 WU7,uqE]j,쀡 Xt={vNcHby1U(ޏJuG)I|Zr'wJ|Jco⚣>Y@ hlwg(qF|۰YUI_Hi ]9h9kzOΣńfo!4+j26 X&l@z)g:JFXW<4|z h ]<]+XCѳ,-NJ Mn6R2ә7%Pߗ#P b4EWG3{d 0!Z!.ȇkcGN4d`vɫ4B' :Vk9 ) sqN(¼#ԭ%cMTRw7%) 7בm6&RHyjLgy<ՔY)MD/RHyNਯ !8>hO-bW(w48 4`nϦ=\秏ۂvw}mK`05|ئ `ˆ)270 a !m+Z | Fnp1R>3jJk1 :9kXߨPkDẄX%b"{sdf\[e\RV>K@Bw_3E $9!#yQt\n{;5I9oSFo\(і\A)g7Ң/~^CW496ߜnjřDoE G~밆2 }v3x8 _@v8- D{tGj ~a/E 6]CWn<n EHOue>GfGK$[*Up?!#o$/ 1bǼD1bY|+q)oLwg(Bj5T~sh9,VL 0^&Pb-N}9/Ì5<,BK}D._uumT;scji?Ѝ o"v1XΧtRs>}%*|,ۤ1xɢ/ڈB3Moa)QX>>lPx+|hD=A(oF/98GH8^R"xn8T5#0eRW."PȽ6,,uY^Y!k@ f|%y k4Om7d>KrCT3`d"ZΈ^Bw<$~=@eVdBL"O ؑeڬb[&@$O+fXhφgb\"eox3s3&eX/ugoTd-Aphc0 QcZ]ۃ&0oI|ˮct'T'qq_116YQ$ּ愴e\p?dMp \/JL{W@wx&[<&Vtk% 2^cݱh=ZZճO+'eW)Z+1ӕxpAcNi|m~^'%!'Fy#3J"˟v%3:kms~)eÆËfo.fu~1Mf,eߖ4rkѭY@ >0e p}ѱI.?ݽR|o9.-N&'/ l9o!kM}{Ӗݻ%Y#VYϓr/ N&41Z*kc eL}\h#IiwBN{{w/0R aӍ<_2 ~EpzY$J%|1fj,4fwz2L?p-z.RV@ T2؃6%@\aA:`G D{55AN E?0M\kFW1^(2E;LʦM-:mpΎs'p$>q//!y+SuQMTiUwv^oCWXy[,ӁB\!M~B-Er7}'aQjP YsƆ庁,qKϚp=^e{e8(.]2P*LxBj;`唹i"OD{CnZ:' XY|̰s4W GmdVbff.+F.ڜDfrC:|9ҥhyB hhHhD#@d)vgo]uDfLE q3VyfNl.ѺpP/5<Ȟk<oV % !FsZ HfcL?NAX9yLm? O]KTE`ZhƱGTO9 &/ջVs2ӛGN Q >/Ş(Or ~;+JtG #I?/ TfvpJ` akE=qw\hP̐)}c [T{*VMwn.\w!rNp oHl)?b}2-e;F/-N߃ n8y[6 -㠅6< XXmx-mG񉜉^& UtuSSڻME0FVjfcJ' =3}e.LǨ_h6\\ᣓ3ߛ]~&J|WZ2*9D':}"Τ3iY+ UUi_ZDrʂ} /* [Z8rϕ_cnap=}+,\P,+^9y{ A1wb4>K )r)Ռ )Mu >f95ڴ\~.v" eNL];?yńU - ƚh+͉8X}=>;} ; !$#D!"O"lze6KL\KV!amаb\"2f~y VU@ $q^Ujl胹cJ=3m,3#:{)Ct7L (,f"#@o*vhE[p U-F~[گj.]!3 H̝˛8|p !~yKd4c<#e ɧ=}Q}FZ/v2$;#Y&T_ߒwؐY$EcKKbIeh/uN]W· dL,N.H|Ƹ8 16y=R{Sr:֜q5t.ˊ6Gz{ʵ_q(. 7g~GMBk=jt"QQc\1$JŌopw~lkn͆svoKx\@Aw e>:>׳E 6.N+k-ى9u鞡:O !y\r{JϦ{*cq3<2.UK#1WS}8%b߽KGrW^:WҸ/n43\L~ͽO;ϲsY=ջW9y8K0,\oVPY4'd7+*=jm}1 S:ג-hcRp 2c:~%8J2JʦX#;BhBqmd&#Fx1Uj)ȝ<6ݽ m=kY]{׍)I, JZşسb͐ mnBB D젇Ϲb,dӜ|kR&qTEjuA6ʬ$Q 4gcNyllۼ#7cN"(-#Pb1si&+td3pCoP0P'i{0e5 ȘC WY⪲)b /b Kn Wp 5s.VCpUR0~Sb9Կ泺5 P:_n_QF劼ݨcpMtLq,\'ȉ"eԯ(4 p!1/:ZÊڱֈp WdoI ~+^火%Xʰm/W I&[Px>ni1j΀yK@0>Y@ɪ#`WYoYXl.26X bP@ .XBVjL| }\ʬXOYA)ߣ ZHi +;nX9)8s}t<<_dS'XL>" Eآf"_l|2ݏKO1 y*XNQ+VH L(|® %aetb?A1tBHQHc94wF҄-+!bny"KTC_oϣO+c!W)6Y"whX,!dmA}su-;KKw㴢:Ek^ﲦk{'?5>8톒qěJj? A/b}K_E Y)N:[(li& п}]ޞ8{ЄENB6CVT)RdN'd_RʜqJMgI2%nZ Ϙ} [QΊn *h >ҤLZD}3!NM4,Q9)˾>>Ţda?A ߌLpL5md;˶vAbE_S$U'brM _FEzx&P)CR(A7ƴ! NZNHREqU(mvhp3<$H.J">=_# Wpoh,WU9BDgo/5ryh-gAAKǖzJfziIK_"tP Wtt9朗#Ss; NؙuM`aыC>ݯf:.E{ʹlpYn;  fˮj,"|m3PbfZʨy:) ) 8ӆhNa*:>wXq-CJsi #'y!>G)eUbNP렵Zg}~$7Bj#H?kpQ|"og<'6rso '>C ҪqHKwW`›']ga7Sנ dWK }=5R70eRd$ޚ#x\b0w62n p Eqtk3)ec)E sTǰ?OQk&& NƧ^3꺱N.<gk tWKXc%b# =o$ש' va}wWtV93nlg;Ϳ#T*7* ؒNzKfw q}e 讆*t$# e1ep})[ɲQ`'u+t9ƽ?xCAoz9WfZc,{v]Ncݪ}e4T=>載,ef(>Aw#);`vۨS[1̷1aO"e͎r쿪D>& BGJΎ5 #}/NC;U& 81l ,Bӣ[n{g/oL\%+uRcXrc}1hAyNjy_N3y)JE9(?мwNJ`] Y#'S4vmAL`v<1EwEq@)*,m/V[ܺT&`;%ʍroܞ<9O/"0:mx)ͼ[^ںm38i֮I;DrLRJRO^1C8Mp|KFFEzrYwؠ O|.B;#!qc&/{ GLNxٱ "{=&b I++1,|kJ <lEWI_uPh!)`†24@DNd2v6[hc X%ܽdޱqPtZx7¸ژdR} 5·7LQm:D.{֨PG@I2ؚ)!ʹVcq$"4g@"V566]rX \ ZCjrC6~Y aŝ+v )ZWlƽE$hڙbIB$0/DXkNjaA{bUASK|\&eٸ @ͪwIA/mYΪp k"m\ÃHL#-͞冢4pV8qm[PBHS1l*ǣpZC1кKT^ .-20X 7v1>\łiWW5&⴪ޟi' XzR|Oν}} GL` %{ݨfPʚ47mL8"aC00RS-ӕ*B%@Z8V@I'ځW_ 7W ʾu&mUeJ '|uSIeL YT*+(8H3oww]A*7Y$h{)m O&", `%3,'*mX=6VYC㵃4TD~we1S]]+]鍚}VpwMŻf:2w=Ym.3PRi^ڢ[Ặe>~GçW ƙ屗\e"kSˢTb9|d zv.f`0_$wc)[yI (#&h҇09ՠ22d O8wBp܀K{Ń18JZOc-r2" .7u\-&ዐLzGVL(C1*;MNLӪ⾋b=Hf?[uOtN=X\b!!o[\m/5 ҞUKI6)PU<*lAsQ֫79&6u ٓeGGZ:9ĄlHSZTlKy4PۯBе_'2a7a٧6רv Qŕ$xb,CfGٰ/*BןiG]$ٶ\Q @z"6J8 `93E31oTf Jw :,3`dO^(ed1C0duG^_*Y+CEpr\Ã- 0'd'W0/MFu3j Л8i?zjyfԖ Ika^s&EU7-Zd~ =([7qbw!Xsb|Lv'?tr9K#zDJ}ǂRĝv,bE? L 鼒 eεk^e@e d-/'>H)@,{),[YތgqRLKhVr45 /XLLEfߟo<0g_ƣ6Z|:ZTG;?:tu MV͸<Z&x4գR#t8 d8B9l YhN]o-KR%!~5=@p:8{2\B;&w9"bx-L1yeTa^f{8?!U3r*Zxw:fݭd dtKqrG&Q1% OgT~ WL]?==\%y3KHԂl5c87yx25*UJ}7z!*aL``Ȭ#z&#O{KÙ@hD?#<6}RfJvq➏|oes{XbtC?AXksޟiALgelk6qFώ3 |QkI-Э32=q6ȥ}S2K+rL7=ұ4y'c$έ[&{ŷcSQڻLȰ}2u75֖*yAf^G y>m#8*Xs8G448[KQ3.l4O?Bkv[hQB'!s(tC+%pmqLyC[C$2DŽc :@3.=ZYu .=,9~Z~MC)A <|P!%ǮbgG{bXg7;R/:<x\ :*p"mټv,]~0Z=~Y5Nݩ'jSD!1Ihɗ{?Z\XLj %Do%t+@W(ՈhO]ڡ~=Y-3qb 0q pȅJEb6HT_K+er?$u,dJ;"IPqq \ 2qu yQ%W*%kf)x:AHRN\o` = qӇ% qI@| ;W7q`#Ί ixۚ*a{P3dTYOi&j`'\КW3,$ž"ӌP t FTQC&xɇh HK|/2o)| Tqg i+Hő'67eB7lt/κ-%jz5|TP?8%2ЮzإD'/n!P⑵oώx ̱oÛyC־FI|G"s >h I2$F,߽%.l ,^m`O5X+NRA~g8Bszd|!9pX}"zr`~؇a[[-8P\LhaoHo06`|5m iPJk)Ц~ñik@",`g <ˍ m 2]9ɐ8Ro>=%ECQ>Gi)HGs4㖂KZ` F.rI <]cpx?)[>v%1?JuEYSժ@um<_?澱<'O#}1 4Y;)q Y8k(Wc jW9i^V^ f!2ԭ"S q*<]k0O-L݅ 3C[d$$,0> p_T|៻~U pbzNx-|/rE 8@Cv[(6F4n.s9D ^r윣(qɄ/vu ֟' cȏI_ yxU1^im/M4-ζZFK  jNTV%H}T*ǧH# 4krɑ ZAp ${"1 'Jx8} Ekm<*z71i-x IF JXT4+5!z#RsI{͍~ }-9[CoRw~ٜm#39ૢfd7y usR4~T`BVZpSixI9?lDFUhًwaK AwVܖz:TO~iᆐh },UX/?3k&x:W6?-pQnAxFC TǣNo˛Ja(oٚU1zsA~LY(w jX fhKpd @W&UwQ}+qbpo֤1h3xGꈚPəę^^s`W` 劬@wC8Pͨ_NY͜q shM0XCOIY Es0s'MY < 0PP6) iuOQW#2ZIgJP6 gզu8|%i5L5ʆUljϕwxh3-"9M} `qBfOm],TxƵJDԾq.Zz]S_Y7KUPr\[6^oghm-xe. #Fߝ +P& M}s1b qF_yv9Nbplڃ3}ȲA9=կz G s8Lȇ%/!klXVŠ+C͙x CM08A3;4hH0[7+zo|CzL4V'Yk|ІsŒHͯ[!l!psŒ[M`DU\6äD|}"[iٷJniN5}-r)b/q4s 'G߹,Y:+Ed\%ʰN5a]D V}c!S7[W(D[RMZ^v+* *Ry 701CKyDU;cyA)1TDkzuIw_/3Gj 9'&wl6Y%5N|sRӃ#?_bL _FQ!_ ᑟG#K: 7W|hM#s֔NmB!isQMR;@! lɘ`q֡odӶ(avTZ0ҘTR57Ls} "rF՗AQjB`F6c'Rˬdf3! P#!\Ta`1Ѭ QfK Lr;+1YSWjAbm*e!mXw׏؊~qodu&m5!sMq*܏@~#NͫT kIF"}. ͗w)  LF(ZR{IoG쾵1K4X h؉:V%)N &->|Su۷0؋?H&SԲ؞!Y2$>TŸ2>^1SPK)hw+]XZ5c S@3b d"\dėjaJssH ƍ E1_+nJ,\Ƃ=KadǧJ^{xOܻ5CXQx:MP%()7Ss14?){:)v1AlDg_U7y]"t ]bIHsΎyƷIVl@tbGHx~M)x^C 7agQ^U3Y=+8ݔ7s,+W>Wi0=dH83#ճk&YQ{Q#.D`A0f;Ͼ+ذ-(ѯp QiD'5.g8UwC䘾rhT[`lj2҃C<g1u9ROM2WNއ6Cɺ>DO)e%e0 O (;~.;'6S(V|8 Ӹc`a z YZ