VSM: A Versatile Semi-supervised Model for Multi-modal Cell Instance SegmentationDownload PDF

29 Nov 2022 (modified: 10 Mar 2023)Submitted to NeurIPS CellSeg 2022Readers: Everyone
Abstract: Cell instance segmentation is a fundamental task in analyzing microscopy images, with applications in computer-aided biomedical research. In recent years, deep learning techniques have been widely used in this field. However, existing methods exhibit inadequate generalization ability towards multi-modal cellular images and require a considerable amount of manually labeled data for training. To overcome these limitations, we present VSM, a \underline{v}ersatile \underline{s}emi-supervised \underline{m}odel for multi-modal cell instance segmentation. Our method delivers high accuracy and efficiency, as verified through comprehensive experiments. Additionally, VSM achieved a top-five ranking in the Weakly Supervised Cell Segmentation category of the multi-modal High-Resolution Microscopy competition.
6 Replies