Object Localization noteebok created as a part of Flipkart GRiD contest 2019
The model was trained for 200 epochs using google colaboratory and achieved an accuracy of 91%. This was the initial model built on top of reset-34 architecture. Further improvement in accuracy was achieved using data augmentation and hyperparameter tunning on the CNN.
Download the image dataset: https://drive.google.com/drive/folders/1FbyOAkNoR0mZsaj6wj93UYtmL5zOSCq9?usp=sharing
Download the train csv: https://drive.google.com/file/d/1jTod5rZ4jzfhx1RDCK2Rc3g6w3SkS02Y/view?usp=sharing