Before Diving deep into Dense Annotation, Let me go through Mask RCNN
The project cites to the following papers on paper directory. And cites following repo
- SSD: Single-Shot MultiBox Detector
- Mask R-CNN for Object Detection and Segmentation
- Instance Segmentation i.e object Detection and Semantic Segmentation
- uses Faster Region Convolution Neural Network (FCNN) and Fully Connected Network (FCN)
However frame rate is too low. Requires improvement on the algorithm, also my machine doesn't have an GPU