End-to-end multilingual recognition on MLT 2019 (test)
58.2F-measure (%)CRAFTS
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CRAFTSsource=original paper [3]2021.03 | 58.2 | 72.9 | 48.5 | |
| CRAFTS (paper)Source=Original paper reported results2023.05 | 58.2 | 72.9 | 48.5 | |
| CRAFTSsource=official ICDAR19-MLT website2021.03 | 51.7 | 65.7 | 42.7 | |
| CRAFTS2023.05 | 51.7 | 65.7 | 42.7 | |
| DeepSolo++Backbone=Res-50, Routing=True, Evaluation identifier=#32023.05 | 51.2 | 62.3 | 43.5 | |
| DeepSolo++Backbone=Res-50, Routing=True, Evaluation identifier=#12023.05 | 50.3 | 62.5 | 42 | |
| Grouped TextSpotterNumber of heads=52023.05 | 48.5 | 67.7 | 37.8 | |
| Multiplexed TextSpotter2021.03 | 48.2 | 68 | 37.3 | |
| Multiplexed TextSpotter2023.05 | 48.2 | 68 | 37.3 | |
| Single-head TextSpotter2021.03 | 39.7 | 71.8 | 27.4 | |
| Single-head TextSpotter2023.05 | 39.7 | 71.8 | 27.4 | |
| RRPN+CLTDR2021.03 | 33.8 | 38.6 | 30.1 | |
| RRPN+CLTDR2023.05 | 33.8 | 38.6 | 30.1 | |
| E2E-MLT2021.03 | 26.5 | 37.4 | 20.5 | |
| E2E-MLT2023.05 | 26.5 | 37.4 | 20.5 |