Research map / Detection and segmentation / CNN features
Segmenter: Transformer for Semantic Segmentation
Robin Strudel, Ricardo Garcia, Ivan Laptev, Cordelia Schmid
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 |
|---|---|---|---|
| add stochastic depth 0.1 (ablation) | Seg-S/16 · ADE20K val | mIoU: 45.01 -> 45.37 | Improved |
| add dropout 0.1 (ablation) | Seg-S/16 · ADE20K val | mIoU: 45.01 -> 42.02 | Hurt |
| replace linear decoder with mask transformer (ablation) | Seg-B†/16 · ADE20K val | mIoU: 47.10 -> 48.70 | Improved |
| replace linear decoder with mask transformer (ablation) | Seg-B/32 · ADE20K val | mIoU: 43.07 -> 44.19 | Improved |
| replace linear decoder with mask transformer (ablation) | Seg-L/16 · ADE20K val | mIoU: 50.71 -> 51.30 | Improved |
| replace linear decoder with mask transformer (ablation) | Seg-L/16 · Pascal Context val | mIoU: 56.5 -> 59.0 | Improved |
| reduce patch size from 32 to 16 (ablation) | Seg-B · ADE20K val | mIoU: 43.07 -> 48.06 | Improved |
| reduce patch size from 16 to 8 (ablation) | Seg-B · ADE20K val | mIoU: 48.06 -> 49.54 | Improved |
| double token dimension (ViT-S to ViT-B) (ablation) | Seg/16 · ADE20K val | mIoU: 45.37 -> 48.06 | Improved |
| double number of layers (ViT-B to ViT-L) (ablation) | Seg/16 · ADE20K val | mIoU: 48.06 -> 50.71 | Improved |
| increase training set from 4k to 20k images (ablation) | Seg-S/16 · ADE20K val | mIoU: 38.31 -> 45.37 | Improved |
| add ImageNet-21k pre-training (SGD) (ablation) | Seg-S/16 · ADE20K val | mIoU: 12.51 -> 45.37 | Improved |
Closest papers by meaning
- FeedFormer: Revisiting Transformer Decoder for Efficient Semantic Segmentation (AAAI 2023 · similarity 0.93)
- Rethinking Semantic Segmentation From a Sequence-to-Sequence Perspective With Transformers (CVPR 2021 · similarity 0.93)
- Rethinking Decoders for Transformer-based Semantic Segmentation: A Compression Perspective (NeurIPS 2024 · similarity 0.93)
- A Transformer-Based Decoder for Semantic Segmentation with Multi-level Context Mining (ECCV 2022 · similarity 0.92)
- Head-Free Lightweight Semantic Segmentation with Linear Transformer (AAAI 2023 · similarity 0.91)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · 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.
