ComputerVisionToolkit

My personal cross-platform toolkit for doing Image Processing and Intelligent Video Analytics experiments. It comes with:

• lib.cvtoolkit library encapsulating the most common and helpful functions

• a set of samples showing how to implement various computer vision tasks

  Samples list

  
  

  • Background subtraction (exponential forgetting, KNN, MOG2)
  • Illumination estimation
  • Image processing (color filters, image derivatives, smoothing, histograms)
  • Image classification (InceptionV3, EfficientNet)
  • Semantic segmentation (Mask R-CNN)
  • Object detection (YOLO)
  • Monodepth
  • Motion detector
  • Optical flow
  • Shadow removal

  
  
  

Requirements

• GCC 9.3.0 compiler for Linux / VS 2019 for Windows
• CMake >= 3.18
• OpenCV >= 3.0:
  • (optional) with CUDA support
• (optional) LibTorch >= 1.8.2
• (optional) ONNX Runtime >= 1.10.0

Build & Install

With CMake

The whole project can be built with CMake

export OpenCV_DIR=path/to/opencv
# (optional) export ENABLE_OPENCV_CUDA=ON
# (optional) export Torch_DIR=path/to/libtorch
# (optional) export Onnxruntime_DIR=path/to/onnxruntime

mkdir build && cd build
cmake ..
make
make install

With VS Code

Add this to .vscode/settings.json

"cmake.configureSettings": 
{
    "OpenCV_DIR": "path/to/OpenCVConfig.cmake",
    "Torch_DIR": "path/to/TorchConfig.cmake",
    "Onnxruntime_DIR": "path/to/onnxruntime",
}

Run

Choose desired task in bin/ directory (like laplacian) and just execute it.

cd bin
./laplacian -?