Robust Vision Transformer: Patch Embeddings & RVT Blocks

This lecture presents the final project presentations of CS-503 on Visual Intelligence, focusing on bio-inspired neuromorphic robotic navigation, graph convolutional networks, and the performance comparison of different architectures for robust vision transformers. The students propose a Ring Attractor Network for navigation, fine-tuning pre-trained networks, and evaluating the results against the baseline. They discuss related work on ring attractors, present results on different architectures, and conclude with limitations and possible extensions for future work.