Research map / Detection and segmentation / Open-set and few-shot detection
Multi-scenario Overlapping Text Segmentation with Depth Awareness
Yang Liu, Xudong Xie, Yuliang Liu, Xiang Bai
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 HSOT synthetic pre-training (ablation) | Mask2Former · MOT test | IoU Ov: 52.82 -> 59.56 | Improved |
| add HSOT synthetic pre-training (ablation) | Mask2Former · MOT test | mIoU Text: 70.28 -> 74.18 | Improved |
| add depth-guided decoder (ablation) | Mask2Former + HSOT · MOT test | mIoU Text: 74.18 -> 75.53 | Improved |
| add depth-guided decoder (ablation) | Mask2Former + HSOT · MOT test | IoU Occd: 76.98 -> 79.17 | Improved |
| add depth-guided decoder without pre-training (ablation) | Mask2Former · MOT test | mIoU Text: 70.28 -> 72.55 | Improved |
| increase synthetic pre-training data from 140k to 280k (ablation) | Ours pre-train · MOT test | mIoU Text: 74.86 -> 75.53 | Improved |
| increase synthetic pre-training data from 280k to 350k (ablation) | Ours pre-train · MOT test | mIoU Text: 75.53 -> 75.59 | No clear effect |
| add scene-related synthetic data only (ablation) | Ours (real data only) · MOT test | mIoU Text: 72.55 -> 73.24 | Improved |
| add document-related synthetic data only (ablation) | Ours (real data only) · MOT test | mIoU Text: 72.55 -> 75.13 | Improved |
| combine document and scene synthetic data (ablation) | pre-training on one data type · MOT test | mIoU Text: 75.13 -> 75.53 | Improved |
| segment overlapping text before recognition (ablation) | SVTRv2 recognizer · MOT test subset | recognition accuracy: 4.04 -> 78.15 | Improved |
| use depth-guided model vs MFM | MFM · SignaTR6K | PT IoU: 73.05 -> 90.46 | 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.
- EAFormer — EAFormer: Scene Text Segmentation with Edge-Aware Transformers (ECCV 2024) · better on 1 of 1 shared result
- Segformer — SegFormer: A Topic Segmentation Model with Controllable Range of Attention (AAAI 2023) · 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
- Deep Matching Prior Network: Toward Tighter Multi-Oriented Text Detection (CVPR 2017 · similarity 0.87)
- Image-Text Co-Decomposition for Text-Supervised Semantic Segmentation (CVPR 2024 · similarity 0.86)
- Rethinking Text Segmentation: A Novel Dataset and a Text-Specific Refinement Approach (CVPR 2021 · similarity 0.86)
- WAS: Dataset and Methods for Artistic Text Segmentation (ECCV 2024 · similarity 0.85)
- Visual-Text Cross Alignment: Refining the Similarity Score in Vision-Language Models (ICML 2024 · similarity 0.84)
- Towards Unified Multi-granularity Text Detection with Interactive Attention (ICML 2024 · similarity 0.84)
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.
