### SFC: Shared Feature Calibration in Weakly Supervised Semantic Segmentation --- - Shared Feature Calibration (SFC) to alleviate over/under-activation ![image](https://hackmd.io/_uploads/BkqksOaVR.png) --- - Framework ![image](https://hackmd.io/_uploads/ByW1VF64R.png) --- - Classifier weight CAM ![image](https://hackmd.io/_uploads/rk_F4KaEA.png) - Prototype CAM ![image](https://hackmd.io/_uploads/S1soVK6NC.png) ![image](https://hackmd.io/_uploads/ByL6NFTE0.png) --- - Multi-scale Distributed-Weighted consistency loss ![image](https://hackmd.io/_uploads/rkLlHY6NC.png) ![image](https://hackmd.io/_uploads/BkJ4rKpNC.png) --- ### Hunting Attributes: Context Prototype-Aware Learning for Weakly Supervised Semantic Segmentation --- - prototype-aware ![image](https://hackmd.io/_uploads/rJa24ziQC.png) --- - Framework ![image](https://hackmd.io/_uploads/rJLRVzj7R.png) --- - top-K context-aware ptototype tailored to the current instance ![image](https://hackmd.io/_uploads/r1rcJcT40.png) ![image](https://hackmd.io/_uploads/BkQ0Jcp4R.png) ![image](https://hackmd.io/_uploads/HkWxlcaVC.png) --- ---
{"title":"SFC & Hunting Attributes: Context Prototype-Aware Learning for Weakly Supervised Semantic Segmentation","description":"type: slide","contributors":"[{\"id\":\"4724c2c9-74e7-418c-87c3-79a619b97461\",\"add\":1172,\"del\":6}]"}
    139 views