Research map / Video / Video-language understanding
End-to-End Video Text Spotting with Transformer
Weijia Wu, Yuanqiang Cai, Chunhua Shen, Debing Zhang, Ying Fu, Hong Zhou, Ping Luo
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 multi-oriented box matching (cosine loss) (ablation) | TransDETR · ICDAR2015 video detection | F-measure: 61.5 -> 75.0 (+13.5%) | Improved |
| add multi-oriented box matching (cosine loss) (ablation) | TransDETR · ICDAR2015 video tracking | IDF1: 41.1 -> 65.5 (+24.4%) | Improved |
| add multi-oriented box matching (cosine loss) (ablation) | TransDETR · ICDAR2015 video tracking | MOTA: 18.4 -> 47.7 | Improved |
| use cosine angle loss instead of L1 angle loss (ablation) | TransDETR · ICDAR2015 video | IDF1: 63.3 -> 65.5 | Improved |
| use cosine angle loss instead of L1 angle loss (ablation) | TransDETR · ICDAR2015 video | MOTP: 76.1 -> 74.1 | Hurt |
| use query-based temporal tracking loss instead of IoU-based association (ablation) | Deformable-DETR text tracker · ICDAR2015 video | IDF1: 56.3 -> 65.5 | Improved |
| use query-based temporal tracking loss instead of transcription-based association (ablation) | Deformable-DETR text tracker · ICDAR2015 video | IDF1: 56.9 -> 65.5 (+8.6%) | Improved |
| use query-based temporal tracking loss instead of IoU association (ablation) | Deformable-DETR text tracker · ICDAR2015 video | MOTP: 75.8 -> 74.1 | No clear effect |
| increase input video sequence length from 3 to 6 frames (ablation) | TransDETR · ICDAR2015 video tracking | IDF1: 61.2 -> 65.5 (4.3%) | Improved |
| increase input video sequence length from 3 to 6 frames (ablation) | TransDETR · ICDAR2015 video tracking | M-Lost: 34.5 -> 32.1 | Improved |
| increase number of decoder layers up to 6 (ablation) | TransDETR · ICDAR2015 video | accuracy | Improved |
| shrink image shorter side below 500 pixels (ablation) | TransDETR · ICDAR2015 video | accuracy: drops at least 5% | Hurt |
Closest papers by meaning
- GoMatching: A Simple Baseline for Video Text Spotting via Long and Short Term Matching (NeurIPS 2024 · similarity 0.92)
- GLOMA: Global Video Text Spotting with Morphological Association (ICLR 2025 · similarity 0.90)
- VimTS: A Unified Video and Image Text Spotter for Enhancing the Cross-Domain Generalization (TPAMI 2025 · similarity 0.87)
- SwinTextSpotter: Scene Text Spotting via Better Synergy Between Text Detection and Text Recognition (CVPR 2022 · similarity 0.87)
- ESTextSpotter: Towards Better Scene Text Spotting with Explicit Synergy in Transformer (ICCV 2023 · similarity 0.87)
- Bridging the Gap Between End-to-End and Two-Step Text Spotting (CVPR 2024 · 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.
