Team Name - Kaggle Master
Team Members - Shreesh Gupta (CS23MTECH12009), Shrenik Ganguli (CS23MTECH14014)
- We were given a dataset which is a subset of the ImageNet Large Scale Visual Recognition Challenge 2012 dataset (ILSVRC2012)
- Our dataset consists of the train and test sets. The train dataset has 50 classes (instead of 1000) and each class originally had 1300 samples. The test dataset has 38,366 images.
- Our challenge was to accurately predict the classes of the images in the test dataset and achieve highest accuracy on private + public leaderboard on this standard multi class classification task
Achieved this through ResNet-152 with preprocessing (image resizing, data augmentation). Explained in detail in the presentation (Kaggle Presentation.pptx) uploaded.
Final standings achieved by this: Rank-2 on Private Leaderboard, Rank-1 on Public Leaderboard
Public LeaderBoard (Score - 0.90533)
Private LeaderBoard (Score - 0.87098)
