This repository contains the work of the paper which is being written for CVPR 2020.
The description of each script is stated here-
- Augmentation.ipynb - The script is for Augmenting the images which are minority in class.
- CNN+Graph.ipynb - The script is for testing the model by passing the shape and color based features through a FCNN
- CNN_feature_extraction- This script is for extracting the shape based features.
- Graph Feature.ipynb- This script is for extracting graph based features like adjacency matrix,etc.
- GraphNN.ipynb- This script is for passing color based feature through a FCNN
- graph_feature_extraction- This script is for color based feature extraction.
- distribution- This script is for understanding the distribution of the images in the dataset
- embedding- This script is for building the embedding for the graph.
- graph_embedding + nn - This script is for trying out the model of passing the graph embedding through a FCNN
- embedding+graph+nn- This script is the best model which ensembles graph embedding and shape based features and regional adjacency graph.
The Harp implementation is inspired and taken through this repository- https://github.com/GTmac/HARP An official implementation by the author