search

facebookresearch / deepcluster

Deep Clustering for Unsupervised Learning of Visual Features
Other
1.69k stars 325 forks source link

Visualization of the 5th conv layer is a noise? #44

Closed zapaishchykova closed 5 years ago

zapaishchykova commented 5 years ago

Hallo! First, thank you for your amazing work!

Secondly, I am wondering why am I receiving a noise from visualization a 5th conv layer? I am using pretrained AlexNet and imageNet-tiny dataset(10 k pics), first I run main.sh and then gradient_ascent.sh. I tried with default settings as well as tried to change number of training epochs, batch size and nothing really changed - I am still recieving noise in con layer 5. Do you have any suggestions about why it doesn't work?

Also a question about the parameter --nmb_cluster - why is it by default 10k and what is it's meaning?

kynantan commented 5 years ago

I've also experienced images that appear to be only noise when running gradient_ascent.sh on any model (any layer), including the pre-trained AlexNet and VGG16 models with unmodified code. (and thanks very much for providing this work :) )

mathildecaron31 commented 5 years ago

Hi, Thanks for your interest, the --nmb_cluster corresponds to the number of clusters. You can refer to our DeepCluster paper for more details.

mathildecaron31 commented 5 years ago

You should play with the hyper-parameters to find a regime where the visualisations are good. For example with the pre-trained deepcluster AlexNet, for conv1 using a learning rate of 3 and 30.000 iterations works well. For conv5, using a learning rate of 30 and 3.000 iterations gives nice images with the other parameters set to their default values.