This the OCR repository of DAVAR Lab, from Hikvision Research Institute, China.
We begin to maintain this code repository to release the official implementations of our recent academic publishments in OCR.
Note: Due to the policy limits of the company. All of the codes were re-implemented based on the open-source frameworks, mmdetection-1.2.0 and mmcv-0.4.2, from open-mmlab. Therefore, the demonstrated results might be slightly different from the reported performances.
To date, we have released / will release the following algorithms:
-
MANGO (to be released) (AAAI 2021)
-
SPIN (to be released) (AAAI 2021)
-
FREE (to be released) (TIP 2020)
-
TRIE (to be released) (ACM MM 2020)
-
Text Perceptron (AAAI 2020)
-
YORO (to be released) (ACM MM 2019)
| Basic Env | version |
|---|---|
| Python | 3.6 |
| cuda | 10.0 |
| cudnn | 7.6.3 |
| opencv | 3.4.9 |
| pytorch | 1.3.0 |
| torchvision | 0.4.1 |
We keep the main part of mmdetection and mmcv exactly same with the official version. Each algorithm is stored under mmdetection/third_party in a separate directory structure.
To Download the repository and install the mmdetection and mmcv, please follow the instructions:
>>> git clone https://github.com/DAVAR-Lab/OCR.git
>>> cd DAVAR-OCR/
>>> bash setup.shIf you want to run some model, you can uncomment the corresponding importing statement in mmdetection/third_party/__init__.py directly in develop mode.
For example, if you want to use the model of Text Perceptron, you could uncomment the line of from .text_perceptron import * in mmdetection/third_party/__init__.py.
Going to the specifc algorithm's directory to see more details.
This project is released under the Apache 2.0 license
The copyright of corresponding contributions of our third-party modules belongs to Davar-Lab, Hikvision Research Institute, China, and other codes from open source repository follows the original distributive licenses.
See latest news in DAVAR-Lab.