Overview
The goal of this project is to adapt Mask R-CNN to yolo, get acceptable quality mask while maintaining high frame rate. This solution is based on darkflow, it's trained on mscoco and currently we get 88% mask accuracy on 5K validation set. Frame rate is very high comparing to original mask r-cnn, it's 20+ fps.
HOWTO train/run, how it works, etc.
TODO...
But the short story is:
- masknet_train.py - train on mscoco
- predict.py - predict on video file/rtsp stream
- best_weights.hdf5 - best weights so far, i.e. those with 88% accuracy
Please note that in this particular demo fps will not be too high because of video decoding, mask prediction and rendering all happen within a single thread, you'll probably get around 10fps. pure prediction however operates on 20+ fps on nVidia TITAN X.