KeyError: load_weights fails because 'layer_names' is missing in extract_features.py

imatge-upc / activitynet-2016-cvprw

Tools to participate in the ActivityNet Challenge 2016 (NIPSW 2016)

https://imatge-upc.github.io/activitynet-2016-cvprw/

MIT License

195 stars 85 forks source link

KeyError: load_weights fails because 'layer_names' is missing in extract_features.py #21

Closed singhania1408 closed 6 years ago

singhania1408 commented 6 years ago

This error was coming on extract_features.py on line model.load_weights(). I am using tensorflow as backend, keras 2.0.1 and changed input dimension to 16, 112, 112, 3 of 1st layer group for working under tensorflow backend. Please help me out so that i can get correct h5 file for extracting features.

4rshdeep commented 6 years ago

Same problem. Any Updates?

singhania1408 commented 6 years ago

@4rshdeep I am confused how I can generate new weight file corresponding to sports dataset for 3d- CN. Do u have any knowledge regarding this??.. No one from admin side is replying?

albertomontesg commented 6 years ago

Check the Keras version using. In the requirements.txt file it is the version used which is a version modified of Keras 1.0.X. Also take into account that the code was tested and used using the Theano backend and with most recent versions may raise some incompatibilities.