The official code of VisionLAN (ICCV2021). VisionLAN successfully achieves the transformation from two-step to one-step recognition (from Two to One), which adaptively considers both visual and linguistic information in a unified structure without the need of extra language model.

• Release code
• Document for Installation
• Trained models
• Document for testing and training
• Evaluation
• re-organize and clean the parameters

2021/10/9 We upload the code, datasets, and trained models.
2021/10/9 Fix a bug in cfs_LF_1.py.
2021/10/12 Correct the typo in train.py

Python2.7
Colour
LMDB
Pillow
opencv-python
torch==1.3.0
torchvision==0.4.1
editdistance
matplotlib==2.2.5

pip install -r requirements.txt

SynthText We use the tool to crop images from original SynthText dataset, and convert images into LMDB dataset.

MJSynth We use tool to convert images into LMDB dataset. (We only use training set in this implementation)

We have upload these LMDB datasets in RuiKe (password:x6si).

Evaluation datasets, LMDB datasets can be downloaded from BaiduYun (password:fjyy) or RuiKe

IIIT5K Words (IIIT5K)
ICDAR 2013 (IC13)
Street View Text (SVT)
ICDAR 2015 (IC15)
Street View Text-Perspective (SVTP)
CUTE80 (CUTE)

The structure of data directory is

datasets
├── evaluation
│   ├── Sumof6benchmarks
│   ├── CUTE
│   ├── IC13
│   ├── IC15
│   ├── IIIT5K
│   ├── SVT
│   └── SVTP
└── train
    ├── MJSynth
    └── SynthText

Download our trained model in BaiduYun (password: e3kj) or RuiKe (password: cxqi), and put it in output/LA/final.pth.

CUDA_VISIBLE_DEVICES=0 python eval.py

Examples of the visualization of mask_c:

   CUDA_VISIBLE_DEVICES=0 python visualize.py

You can modify the 'mask_id' in cfgs/cfgs_visualize to change the mask position for visualization.

Occlusion Scene Text (OST) dataset is proposed to reflect the ability for recognizing cases with missing visual cues. This dataset is collected from 6 benchmarks (IC13, IC15, IIIT5K, SVT, SVTP and CT) containing 4832 images. Images in this dataset are manually occluded in weak or heavy degree. Weak and heavy degrees mean that we occlude the character using one or two lines. For each image, we randomly choose one degree to only cover one character.

Examples of images in OST dataset:

The LMDB dataset is available in BaiduYun (password:yrrj) or RuiKe (password: vmzr)

4 2080Ti GPUs are used in this implementation.

You can follow this implementation to train your own vision model.

Step 1 (LF_1): We first train the vision model without MLM.

   CUDA_VISIBLE_DEVICES=0,1,2,3 python train_LF_1.py

We provide our trained LF_1 model in BaiduYun (password:avs5) and RuiKe (password:qwzn))

Step 2 (LF_2): We finetune the MLM with vision model.

   CUDA_VISIBLE_DEVICES=0,1,2,3 python train_LF_2.py

Download our trained LF_2 model (BaiduYun) (password:04jg) or RuiKe (password:v67q), and put it in /output/LF_2/LF_2.pth.

Then

   CUDA_VISIBLE_DEVICES=0,1,2,3 python train_LA.py

Tips: In LA process, model with loss (Loss VisionLAN) higher than 0.3 and the training accuracy (Accuracy) lower than 91.0 after the first 200 training iters obains better performance.

1. Mask id randomly generated according to the max length can not well adapt to the occlusion of long text. Thus, evenly sampled mask id can further improve the performance of MLM.
2. Heavier vision model is able to capture more robust linguistic information in our later experiments.

If you find our method useful for your reserach, please cite

 @article{wang2021two,
  title={From Two to One: A New Scene Text Recognizer with Visual Language Modeling Network},
  author={Wang, Yuxin and Xie, Hongtao and Fang, Shancheng and Wang, Jing and Zhu, Shenggao and Zhang, Yongdong},
  journal={ICCV},
  year={2021}
}

Suggestions and discussions are greatly welcome. Please contact the authors by sending email to [email protected]