Model card
CLIP4STR-B.
ResearchunknownUnknown paramsUnknown
CLIP4STR: A Simple Baseline for Scene Text Recognition with Pre-trained Vision-Language Model. Base variant. Exploits CLIP pre-training for robust scene text features. Strong on Union14M benchmark. arXiv 2305.14014.
§ 01 · Benchmarks
Every benchmark CLIP4STR-B has a recorded score for.
| # | Benchmark | Area · Task | Metric | Value | Rank | Date | Source |
|---|---|---|---|---|---|---|---|
| 01 | Union14M | Computer Vision · Scene Text Detection | accuracy | 70.8% | #1 | — | source ↗ |
| 02 | host | Computer Vision · Scene Text Recognition | 1-1-accuracy | 79.8% | #2 | 2023-05-23 | source ↗ |
| 03 | coco-text | Computer Vision · Scene Text Detection | 1-1-accuracy | 81.1% | #3 | 2023-05-23 | source ↗ |
| 04 | ic19-art | Computer Vision · Scene Text Detection | accuracy | 85.8% | #3 | 2023-05-23 | source ↗ |
| 05 | uber-text | Computer Vision · Scene Text Recognition | accuracy | 86.8% | #3 | 2023-05-23 | source ↗ |
| 06 | cute80 | Computer Vision · Scene Text Recognition | accuracy | 99.3% | #4 | 2023-05-23 | source ↗ |
| 07 | wost | Computer Vision · Scene Text Recognition | 1-1-accuracy | 87.0% | #4 | 2023-05-23 | source ↗ |
| 08 | icdar2013 | Computer Vision · Optical Character Recognition | accuracy | 98.3% | #5 | 2023-05-23 | source ↗ |
| 09 | svtp | Computer Vision · Scene Text Recognition | accuracy | 97.2% | #5 | 2023-05-23 | source ↗ |
| 10 | iiit5k | Computer Vision · Scene Text Recognition | accuracy | 99.2% | #6 | 2023-05-23 | source ↗ |
| 11 | svt | Computer Vision · Scene Text Recognition | accuracy | 98.3% | #7 | 2023-05-23 | source ↗ |
| 12 | icdar2015 | Computer Vision · Optical Character Recognition | accuracy | 90.6% | #8 | 2023-05-23 | 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 CLIP4STR-B.
- 2023-05-23· Computer Vision· 11 results
CLIP4STR: A Simple Baseline for Scene Text Recognition with Pre-trained Vision-Language Model
§ 04 · Related models
Other Research models scored on Codesota.
§ 05 · Sources & freshness
Where these numbers come from.
papers-with-code
11
results
arxiv-paper
1
result
11 of 12 rows marked verified.