unizard
commented
6 years ago ContextNet combines a deep network at the small resolution with a shallow network at full resolution to achieve accurate and real-time semantic segmentation.
Architecture
Results
BMVC2018 submission
Institute: Toshiba Research, Cambridge URL: https://arxiv.org/pdf/1805.04554.pdf Keyword: Realtime Segmentation Interest: 2
Summary
We propose ContextNet, a new deep neural network architecture which builds on factorized convolution, network compression, and pyramid representations to produce competitive semantic segmentation in real-time with low memory requirements. We analyze our network in a thorough ablation study and present results on the Cityscapes dataset, achieving 66.1% accuracy at 18.2 frames per second at full (1024×2048) resolution.
한줄요약
Corase&Refine 단계로 나누고, 작은영상에 딥하게(coarse), 큰영상으로 쉘로우하게(Refine)