Crnn_chinese_characters_rec

chinese characters recognition

Descriptions in chinese：https://blog.csdn.net/Sierkinhane/article/details/82857572

(Please update pytorch to 1.1.0 https://pytorch.org/ )

Update

• remove lmdb dataloader

• fix ctc loss

Test

• download model password:t99a
  There are images in test_images file, and you just run it as follow.

• python3 test.py

Train

before traning, you should prepare the dataset of characters.(described in csdn blog)

• python3 preprocessing.py to prepare train.txt and test.txt or download password:w877

• modify the path to images, train.txt and test.txt in crnn_main_v2.py

• python3 crnn_main_v2.py --cuda

3.6 million chinese characters dataset：

Data generator:

Results：

Training(accuray was 97.7% ultimately):

##(This repository is based on https://github.com/meijieru/crnn.pytorch)

If interested in Face detection, you can refer to the other repository:

MTCNN https://github.com/Sierkinhane/mtcnn-pytorch

results:

If interested in Machine Learning, you can refer to the other repository:

https://github.com/Sierkinhane/CS229-ML-Implements

CS229-ML-Implements

Implements of cs229(Machine Leaning taught by Andrew Ng) in python.

CS229 Machine Learning Xmind:

CS229 Machine Learning course and notes:

OpenCourse

cs229-Notes

Examples

Linear Regression

Locally Weighted Regression

Logistic Regression

Softmax Regression

Gaussian Discriminant Analysis

TO BE CONTINUED