Research map / Detection and segmentation / CNN features
Exploring the Capacity of an Orderless Box Discretization Network for Multi-orientation Scene Text Detection
Yuliang Liu, Tong He, Hao Chen, Xinyu Wang, Canjie Luo, Shuaitao Zhang, Chunhua Shen, Lianwen Jin
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 orderless box discretization (OBD) (ablation) | Mask R-CNN baseline · ICDAR 2015 | Hmean: 83.5% -> 85.9% | Improved |
| add re-scoring post-processing (RPP) (ablation) | Baseline + OBD · ICDAR 2015 | Hmean: 85.9% -> 86.5% | Improved |
| use KE branch instead of mask branch (ablation) | OBD model · ICDAR 2015 (no augmentation) | Hmean: 79.4% -> 80.4% | Improved |
| add RPP to KE branch (ablation) | OBD model · ICDAR 2015 (no augmentation) | Hmean: 80.4% -> 81.0% | Improved |
| fine-tune with randomly rotated pseudo samples (inconsistent labels) (ablation) | Ours vs CTD · ICDAR 2015 | Hmean: CTD 74.7% -> 50.1%; Ours 80.4% -> 80.7% | Mixed |
| add random crop augmentation (ablation) | OBD baseline · ReCTS validation | Hmean (%): 79.1 -> 81.0 | Improved |
| add random rotation augmentation (ablation) | OBD baseline · ReCTS validation | Hmean (%): 79.1 -> 79.7 | Improved |
| add color jittering (ablation) | OBD baseline · ReCTS validation | Hmean (%): 79.1 -> 79.3 | No clear effect |
| use large training scale (ablation) | OBD baseline · ReCTS validation | Hmean (%): 79.1 -> 81.9 | Improved |
| add ASPP to the backbone (ablation) | OBD baseline · ReCTS validation | Hmean (%): 79.1 -> 77.0 | Hurt |
| add deformable convolution (C4-1) (ablation) | OBD baseline · ReCTS validation | Hmean (%): 79.1 -> 81.7 | Improved |
| replace ResNet-101 with ResNeXt-152-32x8d-FPN-IN5k (ablation) | OBD baseline · ReCTS validation | Hmean (%): 79.1 -> 81.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Textboxes++ — TextBoxes: A Fast Text Detector with a Single Deep Neural Network (AAAI 2017) · better on 2 of 2 shared results
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 0.89)
- EAST: An Efficient and Accurate Scene Text Detector (CVPR 2017 · similarity 0.88)
- TextSnake: A Flexible Representation for Detecting Text of Arbitrary Shapes (ECCV 2018 · similarity 0.88)
- LRANet: Towards Accurate and Efficient Scene Text Detection with Low-Rank Approximation Network (AAAI 2024 · similarity 0.86)
- Towards Unified Multi-granularity Text Detection with Interactive Attention (ICML 2024 · similarity 0.86)
- Box2Poly: Memory-Efficient Polygon Prediction of Arbitrarily Shaped and Rotated Text (AAAI 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.
