Research map / Vision-language and multimodal / Human and embodied modeling
An Alternative Deep Feature Approach to Line Level Keyword Spotting
George Retsinas, Georgios Louloudis, Nikolaos Stamatopoulos, Giorgos Sfikas, Basilis Gatos
Check 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 |
|---|---|---|---|
| Use a common-space encoder for text-string queries | Proposed line-level KWS · IAMDB2 global | MAP (%): 75.31 | Improved |
| Use text-string queries instead of query images | Proposed line-level KWS · IAMDB1 local | MAP (%): 84.25 -> 88.73 | Improved |
| Use query strings instead of query images | Proposed line-level KWS · IAMDB1 global | MAP (%): 73.16 -> 83.15 | Improved |
| Use temporal pyramidal pooling rather than entire-map max pooling (ablation) | PHOC estimator · IAM word-level QbE-PHOC | MAP (%): 83.27 -> 83.98 | No clear effect |
| Use zoning into 5 segments rather than entire-map max pooling (ablation) | PHOC estimator · IAM word-level QbE-PHOC | MAP (%): 83.27 -> 83.66 | No clear effect |
| Use temporal pyramidal pooling rather than entire-map max pooling (ablation) | PHOC estimator · IAM word-level QbE-cfeat | MAP (%): 81.73 -> 82.15 | No clear effect |
| Use zoning into 5 segments rather than entire-map max pooling (ablation) | PHOC estimator · IAM word-level QbE-cfeat | MAP (%): 81.73 -> 82.24 | No clear effect |
| Use temporal pyramidal pooling rather than entire-map max pooling (ablation) | PHOC estimator · IAM word-level QbS-PHOC | MAP (%): 91.11 -> 93.01 | No clear effect |
| Use zoning into 5 segments rather than entire-map max pooling (ablation) | PHOC estimator · IAM word-level QbS-PHOC | MAP (%): 91.11 -> 92.47 | No clear effect |
| Add ROI-based training with expanded word images (ablation) | Proposed line-level KWS · IAMDB1 local QbS | MAP (%): 87.13 -> 88.73 | Improved |
| Add ROI-based training with expanded word images (ablation) | Proposed line-level KWS · IAMDB1 global QbS | MAP (%): 79.14 -> 83.15 | Improved |
| Add ROI-based training with expanded word images (ablation) | Proposed line-level KWS · IAMDB1 local QbE | MAP (%): 83.01 -> 84.25 | Improved |
Closest papers by meaning
- Efficient Learning-Free Keyword Spotting (TPAMI 2018 · similarity 0.88)
- Zero-shot keyword spotting for visual speech recognition in-the-wild (ECCV 2018 · similarity 0.87)
- Neural Ctrl-F: Segmentation-Free Query-By-String Word Spotting in Handwritten Manuscript Collections (ICCV 2017 · similarity 0.87)
- Handwritten Text Recognition: A Survey (TPAMI 2025 · similarity 0.86)
- General Detection-based Text Line Recognition (NeurIPS 2024 · similarity 0.85)
- Bridging the Gap Between End-to-End and Two-Step Text Spotting (CVPR 2024 · similarity 0.83)
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.
