Research map / Detection and segmentation / CNN features
SOTR: Segmenting Objects With Transformers
Ruohao Guo, Dantong Niu, Liao Qu, Zhenbo Li
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 pure twin transformer instead of original transformer (ablation) | SOTR (ResNet-101-FPN) · COCO test-dev | AP: 37.1 -> 39.7 | Improved |
| add two 3x3 convolutions to the twin attention module (hybrid twin) (ablation) | SOTR pure twin transformer · COCO test-dev | AP: 39.7 -> 40.2 | Improved |
| increase original transformer depth from 6 to 12 (ablation) | SOTR original transformer · COCO test-dev | AP: 36.2 -> 37.1 | Improved |
| increase twin transformer depth from 6 to 12 (ablation) | SOTR hybrid twin · COCO test-dev | AP: 37.6 -> 40.2 | Improved |
| use twin transformer instead of original at depth 12 (memory) (ablation) | SOTR · COCO test-dev | memory: 10696M -> 5492M | Improved |
| substitute P5 of FPN with transformer features in the upsampling module (ablation) | SOTR · COCO test-dev | AP: 38.8 -> 40.2 | Improved |
| also substitute P4 with transformer features (ablation) | SOTR · COCO test-dev | AP: 40.2 -> 39.9; APL 73.0 -> 73.7 | Mixed |
| use dynamic convolution kernels instead of static kernels (ablation) | SOTR with twin transformer · COCO test-dev | AP: 38.6 -> 40.2 | Improved |
| add twin transformer to model with dynamic kernels (ablation) | SOTR · COCO test-dev | AP: 39.7 -> 40.2 | Improved |
| use ResNet-101-FPN instead of ResNet-50-FPN backbone (ablation) | SOTR · COCO test-dev | AP: 37.5 -> 40.2 | Improved |
| compare SOTR against Mask R-CNN with the same ResNet-101-FPN backbone | SOTR vs Mask R-CNN · COCO test-dev | AP: 35.7 -> 40.2 (+2.4% AP) | Improved |
| use SOTR-RT instead of YOLACT-550 for real-time instance segmentation | SOTR-RT-736 · COCO test-dev | AP: 28.2 -> 30.7; FPS 44.1 -> 25.2 | Mixed |
Later papers that compared with this method
1 later accepted paper reports SOTR as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Learning Equivariant Segmentation with Instance-Unique Querying (NeurIPS 2022 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 8 of 9 on most shared results.
- SOLOv2 — SOLOv2: Dynamic and Fast Instance Segmentation (NeurIPS 2020) · better on 2 of 2 shared results
- YOLACT-550 — YOLACT: Real-Time Instance Segmentation (ICCV 2019) · better on 1 of 2 shared results
- PolarMask-600 — PolarMask++: Enhanced Polar Representation for Single-Shot Instance Segmentation and Beyond (TPAMI 2021) · better on 1 of 1 shared result
- BlendMask — BlendMask: Top-Down Meets Bottom-Up for Instance Segmentation (CVPR 2020) · better on 1 of 1 shared result
- PolarMask — PolarMask: Single Shot Instance Segmentation With Polar Representation (CVPR 2020) · better on 1 of 1 shared result
- SOLO — SOLO: Segmenting Objects by Locations (ECCV 2020) · better on 1 of 1 shared result
- TensorMask — TensorMask: A Foundation for Dense Object Segmentation (ICCV 2019) · better on 1 of 1 shared result
- ShapeMask — ShapeMask: Learning to Segment Novel Objects by Refining Shape Priors (ICCV 2019) · 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
- SOIT: Segmenting Objects with Instance-Aware Transformers (AAAI 2022 · similarity 0.90)
- Transformer-Based Visual Segmentation: A Survey (TPAMI 2024 · similarity 0.89)
- A Simple Single-Scale Vision Transformer for Object Detection and Instance Segmentation (ECCV 2022 · similarity 0.86)
- Self-Supervised Learning of Object Parts for Semantic Segmentation (CVPR 2022 · similarity 0.86)
- SOLO: Segmenting Objects by Locations (ECCV 2020 · similarity 0.86)
- FeedFormer: Revisiting Transformer Decoder for Efficient Semantic Segmentation (AAAI 2023 · similarity 0.86)
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.
