研究概況
全景分割(Panoptic Segmentation) 在CVPR 2019 被提出,但一年多來已經引起廣泛關注,CVPR 2020 總計有 8 篇文章。概念:即同時實現對背景的語義分割和前景的實例分割。下圖中,(b)語義分割的結果是相同語義的對象標成相同的顏色,(c)實例分割是對目標前景分個體的像素級標註來,(d)全景分割結合了語義分割和實例分割的結果。
論文列表(8篇)
Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation
Code:https://github.com/bowenc0221/panoptic-deeplab
BANet: Bidirectional Aggregation Network With Occlusion Handling for Panoptic Segmentation
Real-Time Panoptic Segmentation From Dense Detections
Bidirectional Graph Reasoning Network for Panoptic Segmentation
Pixel Consensus Voting for Panoptic Segmentation
Video Panoptic Segmentation(Oral)
Code: https://github.com/mcahny/vps
Learning Instance Occlusion for Panoptic Segmentation
Unifying Training and Inference for Panoptic Segmentation