Prof. Yanxi Liu
1. SymNETs: beyond planar symmetry perception Starting with a question: can we train CNN to mimic human perception of real world symmetry? Our experiments on MS-COCO and other data sets illustrate unexpected results, which lead us to rethink about definition of symmetry, in particular, when one perceives the 3D world on a 2D image 2. Dancing with TURKs or Taiji (Tai Chi) with a Master? Inter-subject movement quantification is challenging especially for long (5min+) motion sequences. We explore computational means for spatiotemporal matching and comparison, leading to movements with an intention.