A deep learning library made in C++ using OpenCV used to create CNN architectures. Requires OpenCV to be pre-installed. OpenCV's UMat data type is used for matrix multiplication which makes use of GPU for faster calculations. Supports AMD GPUs.
- Convolutional Layer (filter_h, filter_w, filter_depth, no_of_filters, stride, padding)
- Max Pooling Layer (filter_size, stride, padding)
- Relu Layer (slope)
- Flatten
- Fully Connected Layer (input_nodes, output_nodes)
- Softmax (Scores, Labels)
- Each layer has its separate file. Include to use that layer.
- Layer details are provided while creating that layers object.
- Check out test.cpp file to learn how to fully use this library from using dataset and getting the final model.
- Udhav Sharma (https://www.github.com/UdhavSharma)
- Shubham Bhatnagar (https://github.com/shubham-bhatnagar)