Project README
This project incorporates matterport/MaskR-CNN. Follow the readme file in this repository to enable usage.
!!!! YOU MUST DOWNLOAD THE 'mask_rcnn_cooc.h5' FILE FROM: https://github.com/matterport/Mask_RCNN/releases/download/v2.0/mask_rcnn_coco.h5 AND PLACE IT IN THE MAIN DIRECTORY (this is the NN parameter file used to compute object detections) !!!!
For inference of Tracking by Segmentation Object Representation:
-
Place all videos files inside a directory of any name and place this directory inside this directory
-
Inside above directory change video_directory variable in generate_segmentations.py to match name of your video directory
-
Run generate_segmentations.py - class refinement and confidence (score) threshold can be adapted in segment_video() function.