default evaluate_image.py gives wrong reasults

kentsommer / keras-inceptionV4

Keras Implementation of Google's Inception-V4 Architecture (includes Keras compatible pre-trained weights)

Apache License 2.0

471 stars 175 forks source link

default evaluate_image.py gives wrong reasults #7

Closed Huxwell closed 7 years ago

Huxwell commented 7 years ago

After downloading the code, and running evaluate_image.py on elephant.jpg (everything default, without changing a single line of code or the image) i get:

Loaded Model Weights! Class is: sandbar, sand bar Certainty is: 0.00110015

The same (sandbar) appears for different elephant images I have found. The ordering is TF, the backend is TF as well.

kentsommer commented 7 years ago

@Huxwell

I am assuming you are using Keras 2? If so please use the keras-2 branch that was recently pushed as the weight file format changed between Keras 1 and 2 so it would classify incorrectly if you were using the wrong version.

Also remove the inception-v4 weight files from ~/.keras/models/ before running evaluate_image.py again to ensure the correct weights for your version are downloaded.