Enterprise Navigator

English

简体中文

A shopsign detection app based on AdvancedEAST and AttentionOCR

• This was one of the projects during the "9th China Software Cup" competition (Second Part).
• Our team won the National First Prize in this competition.

Feature

• Target detection. Specifically, shopsign detection
  • During the competition, we used provided dataset to train our models.
• Text recognition
• Detection evaluation
• Result showcase
• Interactive web UI

Dependency

python>=3.7
tensorflow==1.14.0
tensorflow-gpu==1.14.0
tensorpack==0.9.8
keras==2.1.4
pycocotools
gast==0.2.2

How to run

Install requirements

• Make sure you already have a machine with Nvidia graphics that support CUDA.
• Anaconda is highly recommended.

pip install -r requirements.txt

Install models

• By default, if you don't have models installed inside models folder, it will automatically download our trained models from huggingface.
• You can put your own models into models folder.
  • Put AdvancedEAST models into models/AdvancedEAST.
  • Put AttentionOCR models into models/AttentionOCR.

Launch

python launch.py

• Then you can visit the app using the URL below with a web browser like Google Chrome.

http://localhost:6660

License

MIT License

Reference

China Software Cup

AdvancedEAST

AttentionOCR