MatConvNet
lastmod 2026-11-10created 2016-01-072 edits
MatConvNet is an open-source toolbox for building and training convolutional neural networks (CNNs) in MATLAB. It provides operations such as convolution, pooling, and normalization as MATLAB functions and C++/CUDA code compiled with MEX, enabling GPU-accelerated training, and pretrained ImageNet models can be downloaded and used right away. MEX files must be built with vl_compilenn before use, and it was developed by the Oxford VGG group. Research has since moved to PyTorch and others, and maintenance has largely stopped.