Directly utilize CUDA to speed up the processing of GLCM in Python.
This project is a Python package that provides a CUDA-accelerated implementation of the Gray-Level Co-occurrence Matrix (GLCM) features. The package uses the CuPy library, to bridge CUDA and Python code, enabling usage of the GPU when computing GLCM.
This solution, is at least a few thousands times faster than the Scikit-Image implementation, mainly because of the Python loop. However, the benefits of using a GPU is highly limited by the fact that it’s hard to parallelize the accumulation function, which is the main bottleneck we faced when implementing.