CoSWA-YOLOv12 utilizes Cooperative Scale-adaptive Wasserstein Assignment to improve the detection and segmentation of tiny malaria parasites in microscopy. The method applies Wasserstein distance to small objects while reverting to standard assignment for larger ones to prevent localization errors.
HOW THIS AFFECTS YOU
●
researcherYou can solve the gradient-weakness and label-starvation issues common in tiny object detection.
●
healthThis could improve diagnostic accuracy for early-stage malaria in low-resource settings.