Please, fork this repo first!

Instructions

• Your goal is to implement a model using any python library that will recognize the text from the picture as accurate as possible.

• The dataset contained in data/public_data consist of images with English/Chinese texts and info.json file with all ground truth texts.

• You should implement class located in ocr_model.py. Feel free to write any code you want, but keep in mind that the model should match OCRModel interface, so I will be able to easily test it. You can check your recognition quality on public data with main.py script.

• Your grade would depend on both the accuracy of your model and the quality of your code.

  • To measure your accuracy I will download your fork and run your model on the secret version of dataset (with another, but conceptually similar screenshots).

Good luck! And feel free to text me via email if you have any questions:

[email protected]