Skip to content

SEED

1. Introduction

Paper:

SEED: Semantics Enhanced Encoder-Decoder Framework for Scene Text Recognition

Qiao, Zhi and Zhou, Yu and Yang, Dongbao and Zhou, Yucan and Wang, Weiping

CVPR, 2020

Using MJSynth and SynthText two text recognition datasets for training, and evaluating on IIIT, SVT, IC03, IC13, IC15, SVTP, CUTE datasets, the algorithm reproduction effect is as follows:

Model Backbone ACC config Download link
SEED Aster_Resnet 85.20% configs/rec/rec_resnet_stn_bilstm_att.yml 训练模型

2. Environment

Please refer to "Environment Preparation" to configure the PaddleOCR environment, and refer to "Project Clone"to clone the project code.

3. Model Training / Evaluation / Prediction

Please refer to Text Recognition Tutorial. PaddleOCR modularizes the code, and training different recognition models only requires changing the configuration file.

Training

The SEED model needs to additionally load the language model trained by FastText, and install the fasttext dependencies:

1
python3 -m pip install fasttext==0.9.1

Specifically, after the data preparation is completed, the training can be started. The training command is as follows:

1
2
3
4
5
# Single GPU training (long training period, not recommended)
python3 tools/train.py -c configs/rec/rec_resnet_stn_bilstm_att.yml

# Multi GPU training, specify the gpu number through the --gpus parameter
python3 -m paddle.distributed.launch --gpus '0,1,2,3'  tools/train.py -c rec_resnet_stn_bilstm_att.yml

Evaluation

1
2
# GPU evaluation
python3 -m paddle.distributed.launch --gpus '0' tools/eval.py -c configs/rec/rec_resnet_stn_bilstm_att.yml -o Global.pretrained_model={path/to/weights}/best_accuracy

Prediction

1
2
# The configuration file used for prediction must match the training
python3 tools/infer_rec.py -c configs/rec/rec_resnet_stn_bilstm_att.yml -o Global.pretrained_model={path/to/weights}/best_accuracy Global.infer_img=doc/imgs_words/en/word_1.png

4. Inference and Deployment

4.1 Python Inference

Not support

4.2 C++ Inference

Not support

4.3 Serving

Not support

4.4 More

Not support

5. FAQ

Citation

@inproceedings{qiao2020seed,
  title={Seed: Semantics enhanced encoder-decoder framework for scene text recognition},
  author={Qiao, Zhi and Zhou, Yu and Yang, Dongbao and Zhou, Yucan and Wang, Weiping},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={13528--13537},
  year={2020}
}

Comments