Faster R-CNN (with a RPN to predict bounding boxes) architecture with a Mask R-CNN branch to identify and segment cars and people in a "Car & Pedestrian" image dataset obtained from the orginial COCO dataset.
Used TensorFlow framework for development. Used Google Colaboratory to use GPU while training the model.
Please note:
- Before running the Notebook, add the 'data' directory to Google Drive.