Development of computer vision framework based on deep learning and GPU programming. We designed a convolutional and a fully connected neural network for object recognition in images. The training of the network and the computation of its output run on GPU, programmed with OpenCL and C++. The algorithm developed during a master thesis.
For more details check my blog posts:
https://sergioskar.github.io/Neural_Network_from_scratch/
https://sergioskar.github.io/Neural_Network_from_scratch_part2/