how can i initialize the entire model (feature + classifier), and load all weights (81 layers).

[lyu-joe]

i applied '' network.load_weights(os.path.join(args.resume),by_name=True)'', --ghost_cluster 2 --vlad_cluster 8 with your defeat setting , if i don't specify parameter ''by_name=True'',it will raise error 'ValueError: You are trying to load a weight file containing 80 layers into a model with 81 layers.' so dose it means your pre-trained model 'weights.h5' just contains 80 layers, how can i initialize the entire model (feature + classifier), and load all weights (81 layers).

[julik43]

I don't use load_weights, I use restore, and I don't have any ghost_cluster, I think you confused this repository with the one of VGG group which model has ghost_cluster and vlad_cluster.