Conducted on: 17/10/2025
Masked RCNN, YOLOv1, YOLOv2
- How RoIAlign preserves exact spatial locations through bilinear interpolation.
- Advantage of Decoupling
- How Multi Task Learning allows model to focus on spatial layout?
- Brief overview of techniques such as Feature Pyramid Networks (FPN), Data Distillation for unlabeled data
- Reframing object detection as a single regression problem
- Intuition behind YOLO working better at capturing global context than RCNNs
- Mathematical intuition behind Multi-part loss function of YOLO
- Analysis of systemic improvements in YOLOv2 like introduction of anchor boxes to improve recall
- SimCLR V1
- SimCLR V2
Ayushman Dutta
Third Year Attendees: Mukil M Sir, Harshvardhan Saini Sir, Dilshad Sir, Green Sir
Second Year Attendees: Anab, Arnav, Ritesh, Rajat, Arjav, Abhishek, Ayushman, Sreenandan, Anukul
Second Year: None