Handwritten Text Recognition on IAM LAM (test)
2.7CERHTR-ConvText
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HTR-ConvTextParam.=65.9M2025.12 | 2.7 | 7 | |
| HTR-VTParam.=53.5M2025.12 | 2.8 | 7.4 | |
| OrigamiNet-24Param.=115.3M, re-implemented by LAM=true2025.12 | 3 | 11 | |
| OrigamiNet-12Param.=39.0M, re-implemented by LAM=true2025.12 | 3.1 | 11.2 | |
| OrigamiNet-18Param.=77.1M, re-implemented by LAM=true2025.12 | 3.1 | 11.1 | |
| TrOCRParam.=385.0M, extra training data=true, re-implemented by LAM=true2025.12 | 3.6 | 11.6 | |
| CRNNParam.=18.2M, re-implemented by LAM=true2025.12 | 3.8 | 12.9 | |
| GFCNParam.=1.4M, re-implemented by LAM=true2025.12 | 5.2 | 18.5 | |
| ViT + DropKeyParam.=37M2025.12 | 5.7 | 16.5 | |
| CNN + BLSTMParam.=9.3M2025.12 | 5.8 | 18.4 | |
| DeiTParam.=6M2025.12 | 5.9 | 18.7 | |
| ViTParam.=37M2025.12 | 6.1 | 19.1 | |
| TransformerParam.=54.7M, extra training data=true2025.12 | 10.2 | 22 |