Get wrong result

[LeonLIU08]

I use Tensorflow backend and Keras 2.0.3. When I run this code by: -- python evaluate_image.py without changing anything from the command line. I got: Class is: sandbar, sand bar Certainty is: 0.00110018

When I try with another image, the result is the same. I don't know what the reason of this.

[kentsommer]

@LeonLIU08

Unfortunately, this week is a bit crazy for me with deadlines, but I'll look into this first thing next week.

One just off the top of my head suggestion is to make sure you are using the correct branch. If you run the main branch with Tensorflow >= 1 and Keras >= 2 it won't error but you will get horribly wrong results. Perhaps that is the issue here. If not, I'll check into this as soon as I can next week.

[kentsommer]

@LeonLIU08

I've not been able to reproduce this issue with Keras 2.0.3

My suggestion is to ensure that you are indeed on the keras-2 branch by performing the following:

• git fetch origin keras-2
• git checkout keras-2
  • should see:

    Branch keras-2 set up to track remote branch keras-2 from origin.
    Switched to a new branch 'keras-2'

This will ensure you are using the correct branch and everything should work properly.

[anxinzhou]

The keras1-API in the code doesn't work well with keras-2 enviroment.When I turn back to keras-1,I fix the problem.I think update the code with keras-2 API can also work

[kentsommer]

@xxRanger

I'm not exactly sure what you are saying, but unless you are using the correct "keras-2" branch of this repository you are correct it will not work well with keras-2. To clarify, NO CODE needs to be changed to work with keras-2, just make sure you select the correct branch.

[LeonLIU08]

Dear kentsommer,

I have checked my version of Keras though this command: print keras.version I got 2.0.3 And tensorflow version with 1.1.0-rc1 though "print tf.version".

I still have the problem that I put the different images into the system and I got the same result: Class is: sandbar, sand bar Certainty is: 0.0011005

For the installation of Keras, I simply use "sudo pip install Keras".

Could you please tell me how to check the branch I use is correct? Thank you very much.

Best regards, LeonLIU08

2017-04-27 15:14 GMT+08:00 Kent Sommer notifications@github.com:

Closed #9 https://github.com/kentsommer/keras-inceptionV4/issues/9.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/kentsommer/keras-inceptionV4/issues/9#event-1059888691, or mute the thread https://github.com/notifications/unsubscribe-auth/Aap_4R-d9DBNwyH6JQoLu-wPP0lf3CZjks5r0EBSgaJpZM4NF59N .

[kentsommer]

@LeonLIU08

You can type (in the root of the repository): git branch

This will show you something similar to the following:

kent@kbox ~/Documents/scratch/keras-inceptionV4 $ git branch
* keras-2
  master

If you have an "*" next to master or there is no keras-2 listed you should enter the following:

• git fetch origin keras-2
• git checkout -b keras-2

This will ensure you are tracking from the keras-2 branch of this repository and everything should work for you. I'll re-open this, let me know if this works for you or something is confusing.

[kentsommer]

Closing this due to lack of activity.