Table 3 Results on Total-Text
From: An end-to-end text spotter with text relation networks
Method | Detection | End-to-End | |||
|---|---|---|---|---|---|
| Â | P | R | F | Full | None |
TotalText baseline (Ch’ng and Chan 2017) | 40.0 | 33.0 | 36.0 | - | - |
Textboxes (Liao et al. 2017) | 62.1 | 45.5 | 52.5 | 48.9 | 36.3 |
Mask TextSpotter (Lyu et al. 2018) | 87.0 | 80.2 | 83.4 | 71.8 | 52.9 |
TextDragon (Feng et al. 2019) | 85.6 | 75.7 | 80.3 | 74.8 | 48.8 |
det-baseline | 88.5 | 83.2 | 85.7 | - | - |
e2e-baseline | 90.1 | 84.3 | 87.1 | 71.6 | 57.8 |
model with TRN | 90.8 | 83.1 | 86.7 | 74.8 | 62.0 |