“According to technology media news reporters, the latest research proposes a weakly supervised image semantic segmentation network driven by spatiotemporal contrastive learning, which improves segmentation accuracy by mining temporal and spatial supervised information. The experimental results showed that the average intersection to union ratios on the PASCAL VOC and MS COCO datasets reached 72.7% and 43.6%, respectively, demonstrating the superiority of the proposed method.”
Optics and Precision EngineeringVol. 34, Issue 1, Pages: 150-166(2026)