This tutorial lists the text detection algorithms and text recognition algorithms supported by PaddleOCR, as well as the models and metrics of each algorithm on English public datasets. It is mainly used for algorithm introduction and algorithm performance comparison. For more models on other datasets including Chinese, please refer to PP-OCR v2.0 models list.
PaddleOCR open source text detection algorithms list:
On the ICDAR2015 dataset, the text detection result is as follows:
| Model | Backbone | precision | recall | Hmean | Download link |
|---|---|---|---|---|---|
| EAST | ResNet50_vd | 85.80% | 86.71% | 86.25% | Download link |
| EAST | MobileNetV3 | 79.42% | 80.64% | 80.03% | Download link |
| DB | ResNet50_vd | 86.41% | 78.72% | 82.38% | Download link |
| DB | MobileNetV3 | 77.29% | 73.08% | 75.12% | Download link |
| SAST | ResNet50_vd | 91.39% | 83.77% | 87.42% | Download link |
On Total-Text dataset, the text detection result is as follows:
| Model | Backbone | precision | recall | Hmean | Download link |
|---|---|---|---|---|---|
| SAST | ResNet50_vd | 89.63% | 78.44% | 83.66% | Download link |
Note： Additional data, like icdar2013, icdar2017, COCO-Text, ArT, was added to the model training of SAST. Download English public dataset in organized format used by PaddleOCR from:
- Baidu Drive (download code: 2bpi).
- Google Drive
For the training guide and use of PaddleOCR text detection algorithms, please refer to the document Text detection model training/evaluation/prediction
PaddleOCR open-source text recognition algorithms list:
Refer to DTRB, the training and evaluation result of these above text recognition (using MJSynth and SynthText for training, evaluate on IIIT, SVT, IC03, IC13, IC15, SVTP, CUTE) is as follow:
| Model | Backbone | Avg Accuracy | Module combination | Download link |
|---|---|---|---|---|
| Rosetta | Resnet34_vd | 80.9% | rec_r34_vd_none_none_ctc | Download link |
| Rosetta | MobileNetV3 | 78.05% | rec_mv3_none_none_ctc | Download link |
| CRNN | Resnet34_vd | 82.76% | rec_r34_vd_none_bilstm_ctc | Download link |
| CRNN | MobileNetV3 | 79.97% | rec_mv3_none_bilstm_ctc | Download link |
| StarNet | Resnet34_vd | 84.44% | rec_r34_vd_tps_bilstm_ctc | Download link |
| StarNet | MobileNetV3 | 81.42% | rec_mv3_tps_bilstm_ctc | Download link |
| RARE | MobileNetV3 | 82.5% | rec_mv3_tps_bilstm_att | Download link |
| RARE | Resnet34_vd | 83.6% | rec_r34_vd_tps_bilstm_att | Download link |
| SRN | Resnet50_vd_fpn | 88.52% | rec_r50fpn_vd_none_srn | Download link |
| NRTR | NRTR_MTB | 84.3% | rec_mtb_nrtr | Download link |
Please refer to the document for training guide and use of PaddleOCR
For the training guide and use of PaddleOCR text detection algorithms, please refer to the document Text detection model training/evaluation/prediction. For text recognition algorithms, please refer to Text recognition model training/evaluation/prediction
Except for the PP-OCR series models of the above models, the other models only support inference based on the Python engine. For details, please refer to Inference based on Python prediction engine