Dear All,
In week2 assignment when I tested the model with a download cat picture it showed 0 and “non-cat”. My train accuracy was 99% and test accuracy was 70%.
Index5 and index6 of test set are again wrongly classified. Is it because the test accuracy is 70%?
Thanks,
Sunita
The training dataset that we have here is very small for a problem this difficult. The result is that we get a model that happens to work surprisingly well on the specific dataset here, but does not “generalize” very well to arbitrary inputs.
Here’s a thread which discusses this in a bit more detail and shows some experiments with rearranging the training and test data.