Annotation-Efficient Polyp Segmentation via Active Learning

arXiv:2403.14350v1 Announce Type: new
Abstract: Deep learning-based techniques have proven effective in polyp segmentation tasks when provided with sufficient pixel-wise labeled data. However, the high cost of manual annotation has created a bottleneck for model generalization. To minimize annotation costs, we propose a deep active learning framework for annotation-efficient polyp segmentation. In practice, we measure the uncertainty of each sample by examining the similarity between features masked by the prediction map of the polyp and the background area. Since the …

abstract active learning annotation arxiv cost costs cs.cv data deep learning framework however model generalization pixel practice segmentation tasks type via wise