Research map / Detection and segmentation / CNN features
Lightweight and Progressively-Scalable Networks for Semantic Segmentation
Yiheng Zhang, Ting Yao, Zhaofan Qiu, Tao Mei
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use 3x3 Conv instead of 3x3 SepConv as the block (ablation) | multi-path network · Cityscapes val | mIoU (%): 53.9 -> 66.2 | Improved |
| use 3x3 Conv instead of Inverted Residual block (ablation) | multi-path network · Cityscapes val | mIoU (%): 59.2 -> 66.2 | Improved |
| use 3x3 Conv instead of Bottleneck block (ablation) | multi-path network · Cityscapes val | mIoU (%): 55.7 -> 66.2 | Improved |
| use 3x3 Conv instead of Residual block (ablation) | multi-path network · Cityscapes val | latency (ms): 5.24 -> 5.08 | Improved |
| add Bilateral-B bilinear path interaction (ablation) | multi-path network · Cityscapes val | mIoU (%): 66.9 -> 69.9 | Improved |
| use Bilateral-B instead of Bilateral-A interaction (ablation) | multi-path network · Cityscapes val | latency (ms): 6.48 -> 5.97 | Improved |
| use Direct-A interaction instead of Bilateral-B (ablation) | multi-path network · Cityscapes val | mIoU (%): 69.9 -> 67.9 | Hurt |
| add depth expansion (step 1) to tiny network | LPS-Net tiny network · Cityscapes val | mIoU (%): 24.1 -> 41.4 | Improved |
| expand width in steps 2 and 3 | LPS-Net · Cityscapes val | mIoU (%): 41.4 -> 57.8 | Improved |
| expand resolution to two paths with 1/4 second path (step 6) | LPS-Net · Cityscapes val | mIoU (%): 62.3 -> 64.6 | Improved |
| scale from LPS-Net-S to LPS-Net-M by raising main resolution to 1 | LPS-Net · Cityscapes val | mIoU (%): 66.1 -> 69.5 | Improved |
| adopt LPS-Net-S instead of FasterSeg | BDD100K segmentation · BDD100K val | mIoU (%): 55.1 -> 55.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- FasterSeg — FasterSeg: Searching for Faster Real-time Semantic Segmentation (ICLR 2020) · better on 3 of 3 shared results
- BiSeNetV2 — BiSeNet V2: Bilateral Network with Guided Aggregation for Real-Time Semantic Segmentation (IJCV 2021) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Fully Convolutional Networks for Semantic Segmentation (TPAMI 2016 · similarity 0.92)
- Segmentation-Aware Convolutional Networks Using Local Attention Masks (ICCV 2017 · similarity 0.91)
- DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs (TPAMI 2017 · similarity 0.90)
- FasterSeg: Searching for Faster Real-time Semantic Segmentation (ICLR 2020 · similarity 0.90)
- Temporally Distributed Networks for Fast Video Semantic Segmentation (CVPR 2020 · similarity 0.90)
- Fast Neural Architecture Search of Compact Semantic Segmentation Models via Auxiliary Cells (CVPR 2019 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
