We provide a demo script to run our detector on an input image and visualize the detections, as in minimal_demo.m. By default, this script takes images under demo/data and outputs detections to demo/visual.
Clone this project with the --recursive option so that you have my fork of Matconvnet downloaded as a submodule. Make sure it passes all test cases after compilation. Feel free to refer to my compilation code as in matconvnet/compile.m.
Download WIDER FACE and place its data and annotations under data/widerface, following such structure:
data/widerface/wider_face_train.mat(annotations for training set)data/widerface/WIDER_train(images for training set)
Coming soon.