Motion saliency detection software

This software detects whether motion saliency is present for each frame of a video. In a first step, the dominant motion (that is supposed to correspond to camera motion) is compensated. In a second step, saliency detection is realised with a convolutional neural network. The 4 variants are RFS-Motion2D, RFS-DeepDOM, WS-Motion2D and WS-DeepDOM.

Reference

L. Maczyta, P. Bouthemy, O. Le Meur. CNN-based temporal detection of motion saliency in videos. Pattern Recognition Letters, 2019
https://hal.inria.fr/hal-02345209v1