How to retrieve predicted boxes from a predicted image

[GMSL1706]

Hello,

I am working on license plate recognition and I would like to retrieve the boxes predicted by my model in order to be able to recognize the characters of the license plate as in this example

Can someone help me please ?

[fstroth]

Hey, currently we don't have an integration with any ocr system. You might want to take a look for ocr stuff.

[GMSL1706]

Sorry but I just want to make sure. Is there no way to save the predicted boxes as a list, integer or some other variable that we can use for other things?

[fstroth]

All good, the predictions are just regular python objects so you can just save them as you like (pickle, json, etc.)

[GMSL1706]

Can you show me how to do that ? Because I've tried but I didn't find the way to do that

here is an example of my code

model = model_type.model(backbone=backbone(pretrained=True), num_classes=len(parser.class_map), **extra_args) 

# Selection = 0
state_dict = torch.load("./resnet50_fpn_mstrain_2x.pth")
model.load_state_dict(state_dict)

infer_dl = model_type.infer_dl(valid_ds, batch_size=4, shuffle=False)
preds = model_type.predict_from_dl(model, infer_dl, keep_images=True)

show_preds(preds=preds[3:4], show=True, display_label=True, figsize=(15, 20))

And Here is the output of prediction:

Which code could I use to save the predicted boxes please ?

[FraPochetti]

Here it is

[GMSL1706]

Thank you very much. That help me a lot.

As I have the boxes data, I tried to use rectangle function of opencv lib as here on the same image. But I realize that I have to use the predicted image because there was some processing on the original data (I suspect that the boxes data are for processed images... I could be wrong) and it is difficult for me to work with the original image.

Is there any function that can help to save the predicted image with preds[0] as a numpy array something like that according to the boxes data please ?

[FraPochetti]

Check out end2end_detect inside the inference tutotial for what you are asking

[GMSL1706]

THANK YOU VERY MUCH !!!!!!!!!!!! I AM VERRY HAPPY