This is the source code that generates anisotropic supervoxel for video oversegmentation based on GPU :
We provide the source code that tested on Windows.
=================== Runtime Environment ===================
The project requires: OpenGL 4.2 or later; Eigen library; OpenCV with GPU support.
The GLFW3 is provided in the project, you need to download Eigen to the "includes" dir, and install OpenCV with Cuda support, configure the project by CMakeLists and specify the "OpenCV_DIR" in CMakeLists.txt.
=================== Parameters ===================
Specify the paths of input dataset and the output file in main.cpp.
The program generates supervoxel segmentation results, including txt files and pseudo-color images.
Download the video datasets (SegTrackv2, BuffaloXiph) provided in LIBSVXv4.0: http://www.cs.rochester.edu/~cxu22/d/libsvx/
Cite: If you find our code or paper useful, please consider citing
Xiao Dong, Zhonggui Chen, Yong-Jin Liu, Junfeng Yao, and Xiaohu Guo, “GPU-based Supervoxel Generation with a Novel Anisotropic Metric,” IEEE Transactions on Image Processing, 2021.