Research map / Vision-language and multimodal / Human and embodied modeling
Improving OCR-Based Image Captioning by Incorporating Geometrical Relationship
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 geometrical relationships between OCR tokens and use them in the relation-aware pointer network (ablation) | LSTM-R · TextCaps validation | CIDEr-D: 107.4 -> 109.3 | Helped |
| Use multi-label sigmoid loss so one caption word can match multiple OCR regions (ablation) | LSTM-R · TextCaps validation | CIDEr-D: 105.8 -> 106.8 | Helped |
| Determine caption-word types in advance and encode words according to type (ablation) | LSTM-R · TextCaps validation | CIDEr-D: 94.4 -> 99.0 | Helped |
| Clean caption symbols unless they appear in detected image text (ablation) | LSTM-R · TextCaps validation | CIDEr-D: 91.2 -> 94.4 | Helped |
| Use both Rosetta and Google OCR systems instead of only one (ablation) | LSTM-R · TextCaps validation | CIDEr-D: 94.4 -> 99.0 | Helped |
| Include OCR tokens rather than caption without OCR input | M4C-Captioner · TextCaps validation | CIDEr-D: 35.1 -> 89.6 | Helped |
| Use geometrical-relation-aware OCR selection rather than MMA-SR | TextCaps validation · CIDEr-D | CIDEr-D: 98.0 -> 109.3 | Helped |
Closest papers by meaning
- Relational Distant Supervision for Image Captioning without Image-Text Pairs (AAAI 2024 · similarity 0.87)
- Composing Object Relations and Attributes for Image-Text Matching (CVPR 2024 · similarity 0.87)
- PP-OCRv5: A Specialized 5M-Parameter Model Rivaling Billion-Parameter Vision-Language Models on OCR Tasks (CVPR 2026 · similarity 0.85)
- Benchmarking Large Vision-Language Models via Directed Scene Graph for Comprehensive Image Captioning (CVPR 2025 · similarity 0.85)
- OCR-Reasoning Benchmark: Unveiling the True Capabilities of MLLMs in Complex Text-Rich Image Reasoning (ICLR 2026 · similarity 0.85)
- Relation-aware Hierarchical Prompt for Open-vocabulary Scene Graph Generation (AAAI 2025 · 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.