Semantic Segmentation
Semantic segmentation, the task of assigning a semantic label to each pixel in an image, aims to achieve precise pixel-level scene understanding. Current research emphasizes improving accuracy and efficiency across diverse data modalities (RGB, depth, lidar, hyperspectral, and time series) and challenging conditions (low light, adverse weather, imbalanced datasets), often employing advanced architectures like transformers and diffusion models alongside innovative loss functions and training strategies. This field is crucial for numerous applications, including autonomous driving, medical image analysis, remote sensing, and robotics, driving advancements in both model robustness and interpretability.
Papers
Minimalist and High-Performance Semantic Segmentation with Plain Vision Transformers
Yuanduo Hong, Jue Wang, Weichao Sun, Huihui Pan
Cross-attention Spatio-temporal Context Transformer for Semantic Segmentation of Historical Maps
Sidi Wu, Yizi Chen, Konrad Schindler, Lorenz Hurni
Weakly-Supervised Semantic Segmentation with Image-Level Labels: from Traditional Models to Foundation Models
Zhaozheng Chen, Qianru Sun