A feature refinement module for light-weight semantic segmentation network
This work addresses the challenge of maintaining high segmentation accuracy while reducing computational complexity for real-world semantic segmentation tasks, representing an incremental improvement over existing lightweight network designs.
The paper tackles the accuracy degradation problem in lightweight semantic segmentation networks by proposing a feature refinement module (FRM) that extracts semantics from multi-stage feature maps and captures non-local contextual information using a transformer block. The method achieves 80.4% mIoU on the Cityscapes test set with only 214.82 GFLOPs, demonstrating a promising trade-off between accuracy and computational cost.
Low computational complexity and high segmentation accuracy are both essential to the real-world semantic segmentation tasks. However, to speed up the model inference, most existing approaches tend to design light-weight networks with a very limited number of parameters, leading to a considerable degradation in accuracy due to the decrease of the representation ability of the networks. To solve the problem, this paper proposes a novel semantic segmentation method to improve the capacity of obtaining semantic information for the light-weight network. Specifically, a feature refinement module (FRM) is proposed to extract semantics from multi-stage feature maps generated by the backbone and capture non-local contextual information by utilizing a transformer block. On Cityscapes and Bdd100K datasets, the experimental results demonstrate that the proposed method achieves a promising trade-off between accuracy and computational cost, especially for Cityscapes test set where 80.4% mIoU is achieved and only 214.82 GFLOPs are required.