search

ethanwharris / STAWM

Code for the paper 'A Biologically Inspired Visual Working Memory for Deep Networks'
https://arxiv.org/abs/1901.03665
21 stars 2 forks source link

The mnist.gif image #3

Open ifgovh opened 5 years ago

ifgovh commented 5 years ago

Hi, I am wondering how do you plot the gif images. Mainly, how do you transfer the affine matrix to the red rhomboids in mnist.gif? It would be great if you kindly share the code. Thank you very much!

Cheers, Guozhang

ethanwharris commented 5 years ago

Hello,

The visualising code is in visualise.py. Specifically, the red_sqare method is used to make a red square of a certain side which must then be transformed with the same tansform as the sketch, this is done in the model code when draw_stages is True. The gif was not made with code, instead it's just a series of images that were stiched together with an online tool.

Hope that helps :)

ifgovh commented 5 years ago

Hi, That really helps, thank you! The following question is how to access the images saved in state? When I used callbacks.MostRecent() to save state and loaded it, I only found 5 keys which are torchbearer_version, model, optimizer, history, callback_list. Where is the STAGES? Sorry, I am not familiar with torchbearer and I haven't found anything related in its doc. Although I can save the tensor independently, I am curious how to do it in the framework of torchbearer. Thank you for your help!

Cheers, Guozhang

ethanwharris commented 5 years ago

Hi,

The way to do it in torchbearer would be to use a callback. There's a couple of options, either use the callback StagesGrid in visualise.py or use torchbearers MakeGrid callback or similar. The docs can bne found here or there's an example

Hope that helps :)