- To ensure the high reliability of the system, two sets of OCR recognition models from the big data department and the artificial intelligence department need to be simultaneously accessed.
- Provides manual control of the respective task ratios of the two models.
- During the running process, the recognition rate of the model can be calculated dynamically, and the task can be automatically divided according to the recognition rate.
- Have a cache regardless of the recognition result.
- Stores the recognition result information of all models in the database for analysis of the results of model running.
- Use OCR service through RPC service.
- Use Redis to cache bill recognition results, recognition rate of the models, etc.
- Asynchronous communication between systems using Message Queue.
- Use Mysql to store the Recognition result infomations of all models.
- Data persistence layer management with mybatis.
- Use thread pool technology to support high concurrency.