OCR-based Text Recognition on MARIO-7M (test)
61.35APTextDiffuser-10M
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| TextDiffuser-10MPre-defined Text Layout=true, Training Data Size=10M, Training Data Source=Expanded (MARIO-7M, MARIO-TMDB, MARIO-OpenLibrary)2024.06 | 61.35 | 42.89 | 46.83 | 34.25 | |
| SceneTextGen-7MPre-defined Text Layout=false, Training Data Size=7M, Training Data Source=MARIO-7M2024.06 | 52.74 | 44.2 | 44.24 | 30.88 | |
| ControlNetPre-defined Text Layout=true2024.06 | 51.02 | 44.44 | 42.38 | 29.81 | |
| GlyphControl-10MPre-defined Text Layout=true, Training Data Size=10M, Training Data Source=LAION-Glyph2024.06 | 50.75 | 41.18 | 41.4 | 28.83 | |
| TextDiffuser-7MPre-defined Text Layout=true, Training Data Size=7M, Training Data Source=MARIO-7M2024.06 | 47.78 | 34.47 | 36.82 | 25.12 | |
| DeepFloydPre-defined Text Layout=false2024.06 | 24.67 | 27.88 | 22.06 | 13.66 | |
| Latent Diffusion ModelPre-defined Text Layout=false2024.06 | 14.82 | 16.9 | 12.96 | 7.53 |