Brief

Researchers have developed MoEIoU, a novel bounding-box regression loss function for object detection that utilizes a mixture-of-experts approach. This method adaptively combines overlap, center alignment, and aspect-ratio mismatch, with a curriculum-based weighting schedule that prioritizes different error types at various training stages. MoEIoU has demonstrated improved convergence and localization accuracy across multiple datasets and YOLO architectures, outperforming existing state-of-the-art losses. AI

Researchers have developed a new method for detecting defects on high-resolution printed circuit boards (PCBs) that addresses issues with scale and tile boundaries. Their approach involves training detectors on tile crops rather than full-board images to preserve detail, significantly improving detection accuracy. Additionally, a post-processing technique called Topology-Aware Tile Merging (TA-TM) was introduced to reconcile detections across tile edges, enhancing recall for small defects and overall performance without requiring retraining. AI