predict_generator gives "Found 0 images belonging to 0 classes."

Question

I have trained my model on the very big dataset (approx 1200000 train images, and 15000 classes) using Keras' flow_from_directory method. I have successfully trained my model using fit_generator. But, now I am facing the problem for predict_generator.

I got the following error while executing predict_generator:

"Found 0 images belonging to 0 classes."

However, I have converted my train image dataset into the subdirectory based on class labels. But, for test dataset, I don't have any labels. So how can I overcome this issue?

Answer 1

Make a folder Structure.

For instance:

Training_Data
       |
        ---Cat
       |
       -----Dog

Because keras expect your folder in catag