Project Information
- Category: Deep Learning
- Course: NYU - Deep Learning (Yann LeCun and Alfredo Canziani)
- Project date: Spring 2021
- GitHub URL: https://github.com/luke0zong/ssl-classification-movo-v2
Self-Supervised Learning Image Classification with MoCo-v2
This is the final competition for Deep Learning Spring 2021.
The purpose of this project is to train a self-supervised learning model to classify 512,000 unlabeled images (96x96) using 25,600 labeled training images (32 examples, 800 classes).
We adopted the MoCo v2 algorithm and trained the model on NYU Greene HPC (T4 GPU) with PyTorch. The final submission model had an accuracy of 15.98% on validation set.
More information about the project can be found on the GitHub repository.