Nov 28, 2022
Precise and accurate segmentation over boundary areas is important in semantic segmentation. The commonly used convolutional operators tend to smooth and blur local detail cues, making it difficult for deep learning models to generate accurate boundary predictions. In this paper, we propose an efficient boundary-aware convolution operator to boost the boundary modeling capacity for semantic segmentation, named Semantic Difference Convolution (SDC). The SDC is sensitive to the inter-class boundary, while ignoring the noisy intra-class pseudo-boundaries. Based on the SDC operator, we further design a lightweight module, termed Semantic Difference Module (SDM) to enhance the boundary-related information. The SDM can be flexibly plugged into any existing encoder-decoder segmentation model. Extensive experiments show that our approach can achieve consistent improvements (especially for boundary regions) over several typical state-of-the-art segmentation baseline models on four challenging benchmarks, including ADE20K, Cityscapes, COCO-Stuff, and PASCAL-Context.Precise and accurate segmentation over boundary areas is important in semantic segmentation. The commonly used convolutional operators tend to smooth and blur local detail cues, making it difficult for deep learning models to generate accurate boundary predictions. In this paper, we propose an efficient boundary-aware convolution operator to boost the boundary modeling capacity for semantic segmentation, named Semantic Difference Convolution (SDC). The SDC is sensitive to the inter-class boundar…
Account · 952 followers
Professional recording and live streaming, delivered globally.
Presentations on similar topic, category or speaker
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%