This is a basic project to demonstrate the feasibility of the idea, which combining color and depth images to enhance semantic segmentation.
The result is shown in main.ipynb.
The Mask-RCNN part are modified from this project.
wget http://download.tensorflow.org/models/object_detection/mask_rcnn_inception_v2_coco_2018_01_28.tar.gz
tar zxvf mask_rcnn_inception_v2_coco_2018_01_28.tar.gz
Download and extract the needed model files.