Research map / Detection and segmentation / CNN features
FeedFormer: Revisiting Transformer Decoder for Efficient Semantic Segmentation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace the SegFormer-B0 decoder with FeedFormer | MiT-B0 segmentation model · ADE20K val | mIoU: 37.4 -> 39.2 | Helped |
| replace the SegFormer-B0 decoder with FeedFormer | MiT-B0 segmentation model · Cityscapes val | mIoU: 76.2 -> 77.9 | Helped |
| replace the SegFormer-B2 decoder with FeedFormer | MiT-B2 segmentation model · ADE20K val | mIoU: 46.5 -> 48.0 | Helped |
| replace the SegFormer-B2 decoder with FeedFormer | MiT-B2 segmentation model · Cityscapes val | mIoU: 81.0 -> 81.5 | Helped |
| add feature decoding to the model (ablation) | FeedFormer-B0 · ADE20K val | mIoU: 37.1% -> 38.2% | Helped |
| use the FeedFormer head instead of a summation head (ablation) | FeedFormer-B0 · ADE20K val | mIoU: 36.5 -> 39.2 | Helped |
| use the FeedFormer head instead of a concatenation head (ablation) | FeedFormer-B0 · ADE20K val | mIoU: 37.2 -> 39.2 | Helped |
| use the full FeedFormer decoder rather than cross-attention alone (ablation) | FeedFormer-B0 · ADE20K val | mIoU: 37.2 -> 39.2 | Helped |
| use F1, the lowest-level feature, as key and value rather than F2 (ablation) | FeedFormer-B0 decoder · ADE20K val | mIoU: 38.8 -> 39.2 | Helped |
| use F1, the lowest-level feature, as key and value rather than F3 (ablation) | FeedFormer-B0 decoder · ADE20K val | mIoU: 38.1 -> 39.2 | Helped |
| replace SegFormer-B0 with FeedFormer-B0 | MiT-B0 segmentation model · ADE20K val | mBA: 31.2 -> 32.0 | Helped |
Closest papers by meaning
- Rethinking Decoders for Transformer-based Semantic Segmentation: A Compression Perspective (NeurIPS 2024 · similarity 0.91)
- SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers (NeurIPS 2021 · similarity 0.91)
- Head-Free Lightweight Semantic Segmentation with Linear Transformer (AAAI 2023 · similarity 0.90)
- Your ViT is Secretly an Image Segmentation Model (CVPR 2025 · similarity 0.88)
- DECO: Unleashing the Potential of ConvNets for Query-based Detection and Segmentation (ICLR 2025 · similarity 0.87)
- SED: A Simple Encoder-Decoder for Open-Vocabulary Semantic Segmentation (CVPR 2024 · similarity 0.87)
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.