New for: D2, D3
Our latest work aims at enabling more scalable recognition and providing increased robustness in such real-world situations. The talk will give an overview on our progress in learning hierarchical features representations, compensating for domain shifts between training and test data, leveraging different sensor modalities for recognition and detection and how those models can be acquired in a scalable manner.
I'll conclude by highlighting our efforts towards robotic manipulation of cloth and future directions.