Issue with License Plate Text Recognition and Bounding Box Coordinates using GPT API

Hello everyone,

I’m working on a project where I need to recognize the text from vehicle license plates and also determine the bounding box coordinates of the license plates in images. When I use the GPT API, it accurately recognizes the text from the license plates. However, it fails to provide the correct coordinates for the bounding box of the license plate.

The API correctly reads the text on the license plate, but the coordinates it returns for the bounding box are incorrect.

Could anyone explain why this discrepancy occurs and how I can improve the accuracy of the bounding box detection? Are there any specific models, pre-processing steps, or API configurations that you recommend for this type of task?

Thank you for your help!

Hey and welcome to the forum!

This is an expected limitation, you can read more about the limitations here:

I’ll recommend using YOLO for automatic number plate recognition and segmentation instead.

Thank you for the recommendation, I will look into it.

1 Like

Hi Luquepedro,

I’m not answering your question but just out of curiosity, how did you manage correctly recognising licence plates? My implementation recognising plates very differently all the time, accuracy is not that good.

Thanks for any hint :slight_smile: