Attention-calibration based double-branch cross-domain person re-identification

Article Computer Science, Artificial Intelligence

Rectifying Pseudo Label Learning via Uncertainty Estimation for Domain Adaptive Semantic Segmentation

Zhedong Zheng et al.

Summary: This paper proposes an approach to rectify pseudo label learning in unsupervised semantic segmentation adaptation by explicitly estimating prediction uncertainty during training and incorporating uncertainty into the optimization objective. Through extensive experiments, the proposed method achieves significant improvements on three benchmarks.

INTERNATIONAL JOURNAL OF COMPUTER VISION (2021)

Add to Collection

Article Computer Science, Artificial Intelligence

Weakly Supervised Person Re-ID: Differentiable Graphical Learning and a New Benchmark

Guangrun Wang et al.

Summary: In this study, we proposed a method to ease the annotation for person reidentification by grouping images into bags and assigning them with bag-level labels. This led to the creation of a large-scale benchmark called SYSU-30k, which contains 30k individuals. The weakly supervised Re-ID problem was addressed by introducing a differentiable graphical model to generate reliable pseudo labels for learning the Re-ID model.

IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS (2021)

Add to Collection

Article Computer Science, Artificial Intelligence