Curriculum Self-supervised Learning for Weakly-supervised Histopathological Image Segmentation

WSSS4LUAD: Grand Challenge on Weakly-supervised Tissue Semantic Segmentation for Lung Adenocarcinoma

“…Boyd et al [47] introduced a visual field expansion to use the input image to generate the pseudo contents of the surrounding area. Cheng et al [48] proposed a curriculum selfsupervised learning strategy by training several tasks with progressively increasing difficulties. It includes the image reconstruction task, the image inpainting task and the stain deconvolution task.…”