Research map / Detection and segmentation / Open-set and few-shot detection
Using Object Information for Spotting Text
Shitala Prasad, Adams Wai Kin Kong
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 object-label pretraining and object features to the text detector (ablation) | TO-CNN · NTU-UTOI | F-score: 0.59 -> 0.66 | Improved |
| train Faster R-CNN on object and text labels together instead of text only | Faster RCNN · NTU-UTOI | F-score: 0.58 -> 0.37 | Hurt |
| add the third fine-tuning stage of the whole network (ablation) | TO-CNN · NTU-UTOI | F-score: 0.59 -> 0.66 | Improved |
| add the second training stage (Text VGG-16 on frozen object features) (ablation) | TO-CNN · NTU-UTOI | F-score: 0.42 -> 0.59 | Improved |
| use five aspect ratios instead of one with a single scale 8 (ablation) | TO-CNN · NTU-UTOI | F-score: 0.32 -> 0.58 | Improved |
| use three scales instead of one with a single ratio (ablation) | TO-CNN · NTU-UTOI | F-score: 0.32 -> 0.62 | Improved |
| combine three scales with five ratios instead of three scales and one ratio (ablation) | TO-CNN · NTU-UTOI | F-score: 0.62 -> 0.66 | Improved |
| improve anchors for Faster R-CNN pre-trained on NTU-UTOI objects (ablation) | Faster R-CNN · NTU-UTOI | F-score: 0.37 -> 0.57 | Improved |
| pretrain on NTU-UTOI objects instead of COCO objects before fine-tuning on text (ablation) | Faster R-CNN (same anchors) · NTU-UTOI | F-score: 0.24 -> 0.37 | Improved |
| additionally train on COCO-Text text labels | TO-CNN · COCO-Text | F-score: 0.43 -> 0.45 | Improved |
| use object information (TO-CNN) versus prior best on recall | TO-CNN · MSRA-TD500 | recall: 0.76 -> 0.90 | Improved |
Closest papers by meaning
- Towards End-to-End Text Spotting in Natural Scenes (TPAMI 2021 · similarity 0.89)
- Towards Unified Scene Text Spotting Based on Sequence Generation (CVPR 2023 · similarity 0.87)
- Mask TextSpotter: An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes (ECCV 2018 · similarity 0.87)
- Bridging the Gap Between End-to-End and Two-Step Text Spotting (CVPR 2024 · similarity 0.86)
- GoMatching: A Simple Baseline for Video Text Spotting via Long and Short Term Matching (NeurIPS 2024 · similarity 0.86)
- SemiETS: Integrating Spatial and Content Consistencies for Semi-Supervised End-to-end Text Spotting (CVPR 2025 · 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.
