ASFNet: Adaptive multiscale segmentation fusion network for real‐time semantic segmentation. (26th May 2021)

• Record Type: Journal Article
• Authors: Zha, Hengfeng; Liu, Rui; Yang, Xin; Zhou, Dongsheng; Zhang, Qiang; Wei, Xiaopeng
• Is Part Of: Computer animation and virtual worlds. Volume 32:Number 3/4(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-05-26
• Subjects: computer vision -- multiscale fusion -- real‐time semantic segmentation -- segmentation fusion; Computer animation -- Periodicals; Visualization -- Periodicals; 006.6
• DOI: 10.1002/cav.2022
• Languages: English
• ISSNs: 1546-4261

Abstract

Abstract: Recently, the development of deep learning has facilitated continuous progress in the field of computer vision. Pixel‐level semantic segmentation serves as a fundamental task in computer vision. It achieves significant results by connecting wider and deeper backbone networks and building fine‐grained segmentation heads. However, applications such as self‐driving cars are more critical to the computational speed of the algorithms. The trade‐off between accuracy and real‐time performance of existing algorithms is still a challenging task. To address this challenge, this article proposes an adaptive multiscale segmentation fusion network to fuse multiscale contextual, which designs an adaptive multiscale segmentation fusion module based on an attention mechanism. Using segmentation fusion instead of feature fusion, the multiscale segmentation results are aggregated to obtain more precise segmentation results. The final results achieved 70.9% mIoU of accuracy in the Cityspace test set, processing images at 61 FPS when the input is 1024 × 2048. In addition, when adjusting the input size to 512 × 1024, the images are processed at 185 FPS. Abstract : The article proposed a multiscale segmentation fusion module based on the attention mechanism to introduce multiscale contexts for real‐time semantic segmentation, and achieve an efficient and accurate real‐time sematic segmentation network by fusing the output segmentation maps of multiple stages of the backbone network instead of feature fusion. Experiments on cityscapes verify the effectiveness of the proposed method, when the size of input image is 512 × 1024, the proposed method an accuracy of 70.9% mIoU at 185 FPS.