CGNet: A Light-Weight Context Guided Network for Semantic Segmentation

The demand of applying semantic segmentation model on mobile devices has been increasing rapidly. Current state-of-the-art networks have enormous amount parameters hence unsuitable for devices, while other small memory footprint models follow the spirit classification network and ignore inherent characteristic segmentation. To tackle this problem, we propose a novel Context Guided Network (CGNet), which is light-weight efficient We first (CG) block, learns joint feature both local surrounding context effectively efficiently, further improves with global context. Based CG develop CGNet captures contextual information in all stages network. specially tailored to exploit property increase accuracy. Moreover, elaborately designed reduce number save footprint. Under an equivalent parameters, proposed significantly outperforms existing networks. Extensive experiments Cityscapes CamVid datasets verify effectiveness approach. Specifically, without any post-processing multi-scale testing, achieves 64.8% mean IoU less than 0.5 M parameters.