Game-of-Deep-Learning

Ship Classification using Image Processing

This is a multi-class classification problem with 5 types of ships.

It was a 2-week nation-wide competition

Approach

• Applied SOTA practices like Transfer Learning, OpenCV, MixUp etc. on an image classification problem with 5 classes.
• Used architectures like ResNets, NASNets, DenseNets etc
• Used heavy Data Augmentations like Cutout, Colour Dimension shifts etc.
• Generated an ensemble of all the best performing models

Link to my work

https://www.kaggle.com/sandeeppat/ship-classification-top-3-5-kernel

Rank - 25th out of 2000 (Top 3.4 %).