Domain adaption in Object Detection

Project URL:

  • Created a novel architecture with domain adaption components meant to work on minimizing h divergence trained in an adversarial manner.
  • Implemented various efficient transfer learning approaches to prevent over- catastrophic forgetting and facilitate incremental learning.
  • Worked with RPN, SSD, Yolo based models on IDD,KITTI,BDD,cityscapes for real time navigation systems