Mutual learning with discrepancy for weakly supervised object detection

Kequan Yang, Xichen Ye, Yuanchen Wu, Jide Li, Xiaoqiang Li, Pinpin Zhu

Abstract: Highlights•MuDi framework reduces pseudo-label noise in weakly supervised detection.•AFDL method enhances semantic diversity and feature robustness.•Mutual supervision effectively mitigates noise from self-training methods.•Achieves new state-of-the-art on VOC and MSCOCO detection benchmarks.

External IDs:dblp:journals/eswa/YangYWLLZ25