Mask_RCNN using Webcam

Requirements

Python>=3.4
numpy
scipy
Pillow
cython
matplotlib
scikit-image
tensorflow>=1.3.0
keras>=2.0.8
opencv-python
h5py
imgaug
IPython[all]
pycocotools

Model is trained MS COCO dataset and pre-trained weights are used for instance segmentation of the objects using Webcam

Here is the sample video of instance segmentation of objects using Mask-RCNN https://youtu.be/2YfW8VDSvZo

Sample video of this model tested on Busy Indian Roads : https://www.youtube.com/watch?v=o6XQ79yozO8

Sample Images tested using Mask-RCNN

Possible Improvements

This model is implemented using Nvidia 840M with 4GB memory, so the frame rate is very less.For getting better frame rates we can use Nvidia Titan X or Nvidia 1080Ti.By using these graphic cards we can double the frame rate of the present Scenario. Right Now only 80 common objects can be detected, but if we train this on multiple datasets then we can increase the types of objects detetected.