Warnify-Fusion

multimodal prediction of a violent videos (weak label) by a combination of 3 pre-trained models
(I3D - rgb model and optical flow, and audio model)