Counting 3,742 Big Data & Machine Learning Frameworks, Toolsets, and Examples...
Suggestion? Feedback? Tweet @stkim1

Author
Last Commit
Apr. 8, 2019
Created
Sep. 27, 2018

STAWM

Code for the paper 'A Biologically Inspired Visual Working Memory for Deep Networks', originally submitted to ICLR2019, reviews can be found here. The memory module can be found in memory.py and each experiment has its own file.

Examples

MNIST

Reconstruction (sketch) sequences for 36 glimpses / sketches of size 4, memory size: 128, latent size: 4. Note that this is significantly more glimpses than used in the paper in order to obtain a reasonable animation, but also shows that the memory can learn a useful representation over very long sequences.

mnist gif

CelebA

Segmentation over time for the model from the paper. Here we plot the masked images according to the alpha channel of the final sketch. As the model trains this alpha channel moves from a uniform representation, to a segmented facial region.

celeb gif