AttributeError: Can't get attribute 'RCNN' on <module 'network'

[kingstontan]

I'm getting the following error while trying to make predictions.
I fixed it temporarily by changing the class name in network.py from CRNN to RCNN.
Should I be changing the class name?

Also I managed to run the command after this change but my output in predictions.txt is empty.

Thanks a bunch

from CRNN

to RCNN

[melohux]

When we trained the model, we named the network as RCNN. Later we changed it to CRNN. That's where the problem comes from. Thanks a lot, we will fix it very soon. May I take a look at your test image ?

[kingstontan]

Got it. So I should leave it as RCNN for now?

Here is the test image and here's the command I used.

python line_predictor.py --data_path images/ --model_path trained_networks/ICFHR2014_model_imgH64.pth --imgH 89 --save_model_path tesdir/

[melohux]

I have fixed these issues. Please update the code and now use the command line python lines_predictor.py --data_path datapath --model_path ./trained_networks/IAM_model_imgH64.pth --imgH 64 to test the image. Hope it will work well.

[kingstontan]

Hi Hux, I tried the new code with the following command, but the predictions were still empty. Using back the same test images.
python line_predictor.py --data_path iam/ --model_path ./trained_networks/IAM_model_imgH64.pth --imgH 64

The only 2 changes I made were

1. the added map_location=torch.device('cpu') while loading model as I don't have an Nvidia gpu
2. I removed self.time in params.py
   

Thank you so much for your time :)

[melohux]

Please make sure that the folder of data_path contains images in .jpg form.

[kingstontan]

Got it, that was the issue. Thanks for your replies.