Research map / Architectures and efficiency / Attention and transformers
TransMatting: Enhancing Transparent Objects Matting with Transformers
Huanqia Cai, Fanglei Xue, Lele Xu, Lili Guo
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 TGTB tri-token module (ablation) | TransMatting · Composition-1k | SAD: 29.14 -> 27.45 | Improved |
| add MGF multi-scale fusion module (ablation) | TransMatting · Composition-1k | SAD: 29.14 -> 27.21 | Improved |
| add MGF module on top of TGTB (ablation) | TransMatting · Composition-1k | SAD: 27.45 -> 26.83 | Improved |
| add TGTB on top of MGF (ablation) | TransMatting · Composition-1k | MSE: 5.57 -> 5.22 | Improved |
| introduce tri-tokens in all four stages instead of stage 1 only (ablation) | TransMatting · Composition-1k | SAD: 31.68 -> 26.83 | Improved |
| introduce tri-tokens in all four stages instead of stage 4 only (ablation) | TransMatting · Composition-1k | SAD: 29.50 -> 26.83 | Improved |
| add global branch to local branch in MGF (ablation) | TransMatting · Composition-1k | SAD: 27.16 -> 26.83 | Improved |
| add local branch to global branch in MGF (ablation) | TransMatting · Composition-1k | SAD: 27.39 -> 26.83 | Improved |
| add TGTB and MGF on top of MGF-less/TGTB-less base (MSE) (ablation) | TransMatting · Composition-1k | MSE: 6.34 -> 5.22 | Improved |
| replace CNN matting with TransMatting on totally transparent objects | TransMatting vs MGMatting · Composition-1k TT objects | SAD: 77.88 -> 59.37 | Improved |
| train on Composition-1k and test on Transparent-460 (generalization) | TransMatting vs MGMatting · Transparent-460 test | SAD: 344.65 -> 192.36 | Improved |
| use TransMatting instead of FBAMatting with test-time augmentation | TransMatting · Composition-1k test | MSE: 5.2 -> 4.58 | Improved |
Later papers that compared with this method
2 later accepted papers report TransMatting as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Path-Adaptive Matting for Efficient Inference Under Various Computational Cost Constraints (AAAI 2025 · better on 1 of 2 shared results)
- Revisiting Context Aggregation for Image Matting (ICML 2024 · better on 1 of 1 shared result)
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
- Learning Transparent Object Matting (IJCV 2019 · similarity 0.91)
- TOM-Net: Learning Transparent Object Matting From a Single Image (CVPR 2018 · similarity 0.90)
- Semantic Image Matting (CVPR 2021 · similarity 0.89)
- Bridging Composite and Real: Towards End-to-End Deep Image Matting (IJCV 2022 · similarity 0.89)
- Training Matting Models Without Alpha Labels (AAAI 2025 · similarity 0.89)
- DRIP: Unleashing Diffusion Priors for Joint Foreground and Alpha Prediction in Image Matting (NeurIPS 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.
