Project information

  • Category: TB Detection and Classification model and Flask app development
  • Skills employed: HTML, Bootstrap, Javascript, Docker, Python, PyTorch, Flask, MMDetection, MMClassification

TB Detection using Vision-Transformer

  • Trained state-of-the-art vision-transformer based architectures for both the classification and detection tasks.
  • Employed advanced training procedures such as Cutmix and Mixup augmentation, learning rate optimization such Cosine Annealing schedule, and strong regularization techniques such as label smoothing and stochastic depth.
  • Holds first position on the TBX11K Tuberculosis Classification and Detection Challenge leaderboard (https://competitions.codalab.org/competitions/25848#results).