Model card
JSTR.
Fujitakeopen-sourceUnknown paramsDTrOCR + judgment module for image-text matching to reduce misrecognition
JSTR extends DTrOCR with a judgment mechanism that predicts image-text correspondence to improve recognition accuracy. arxiv:2404.05967, Apr 2024.
§ 01 · Benchmarks
Every benchmark JSTR has a recorded score for.
| # | Benchmark | Area · Task | Metric | Value | Rank | Date | Source |
|---|---|---|---|---|---|---|---|
| 01 | icdar-2013 | Computer Vision · Scene Text Detection | accuracy | 99.2% | #1 | 2024-04-09 | source ↗ |
| 02 | icdar2015 | Computer Vision · Optical Character Recognition | accuracy | 98.7% | #1 | 2024-04-09 | source ↗ |
Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.
§ 03 · Papers
1 paper with results for JSTR.
- 2024-04-09· Computer Vision· 2 results
JSTR: Judgment Improves Scene Text Recognition
§ 05 · Sources & freshness
Where these numbers come from.
arxiv
2
results
2 of 2 rows marked verified.