This is an unofficial TensorFlow re-implementation of paper "Efficient and Accurate Arbitrary-Shaped Text Detection with Pixel Aggregation Network" published in ICCV 2019, with Tensorflow 2.
- Backbone model
- FPEM model
- FFM model
- Integrated model
- Loss Function
- Data preprocessing
- Data postprocessing
- Training pipeline
- Inference pipeline
- Evaluation pipeline
python train.py --batch 16 --epoch 601 --dataset_type ctw --gpu True
python inference.py --input ./data/CTW1500/test/text_image --model ./outputs/model_epoch_600 --bbox_type poly
*Default using ResNet-50 backbone model, also support ResNet-101 backbone model.
Model | Precision | Recall | F score |
---|---|---|---|
PAN-320 | 0.8472 | 0.6939 | 0.7629 |
[1] Original paper: https://arxiv.org/abs/1908.05900
[2] Official PyTorch code: https://github.com/whai362/pan_pp.pytorch