[ML] Cats and Dogs Test Set .evaluate() only 1 image?

Python
1

Hi, I have been working on the second ML with Python Projects Cats and Dogs. So far everything seems to work. However, when I reach the point/cell of model.predict() I feel like something is not working correctly.

I trained the model and now I want to grap the test set and get the predictions so I use:

model.evaluate(test_data_gen) - In the output there is only this response 1/1 which I think that it takes only one test image instead of 50?! Not sure about this information.

And I get this output for this cell:

1/1 [==============================] - 0s 232ms/step - loss: 0.0000e+00 - accuracy: 0.0000e+00

[0.0, 0.0]

Also the transformation to a list is missing but I guess the above problem should be resolved first?!

Would be really great to clarify that or any hint would be very nice.

You can find my colab here: Google Colab

2

Compare evaluate and predict to make sure you are using the one you intend. You should get 50 results. Your results should look something like

[[4.86431420e-01 5.13568521e-01]
 [6.01261914e-01 3.98738027e-01]
 [7.98121810e-01 2.01878160e-01]
 [1.52471483e-01 8.47528458e-01]
 [5.39969265e-01 4.60030705e-01]
 [4.93960142e-01 5.06039798e-01]
...
]

if you print them. You’ll need to process this data before doing the comparison against answers in the last cell as it expects binary data.

3

Thank you very much! Looking at the two links made it clear to me and I implemented it then.

I used np.argmax(model.predict(test_data_gen), axis=-1) to get the predicted_classes into binaries and got 66% Accuracy in the end.

Thank you.

4

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.