Due to an increase in the number of inquiries from customers, the front-end has been processing the original pages with OCR to extract text information, which has also significantly increased in frequency. We are using the currently popular ChatGPT and the currently popular neural network inference framework to significantly improve the clarity of images using Real-ESRGAN nucnn Vulkan, greatly improving the accuracy of recognition and accelerating our quotation speed for regular models.
At the same time, using this technology, we are also able to correct and process information from nameplate photos with poor pixel quality so that we can quickly and accurately recognize them without spending more time asking back and forth.
We hope that each step in our technical improvement will bring our customers a better service experience.
