more

Backbone Can Not be Trained at Once: Rolling Back to Pre-trained Network for Person Re-Identification. (AAAI, 2019)  
Skeleton-based Action Recognition of People Handling Objects (WACV, 2019)  
Knowledge Transfer via Distillation of Activation Boundaries Formed by Hidden Neurons (AAAI, 2019). Oral  
Knowledge Distillation with Adversarial Samples Supporting Decision Boundary (AAAI, 2019)  
Pose Transforming Network: Learning to Disentangle Human Posture in Variational Auto-encoded Latent Space (PRL, 2018)  

The goal of the PILab is to gain useful technologies on perception and intelligence which can be applied to visual surveillance systems, inspection machines, robots, etc.

We are developing perceptual primitives to detect, track, and recognize human/vehicles, faces and to understand abnormal behaviors and situations through visual information. In addition, we are developing incremental learning models including probabilistic learning machines for integrating multiple sensory modalities in the changing environments.

more

새소식이 없습니다.
관리자