Jay Taylor's notes

back to listing index


[web search]
Original source (github.com)
Tags: tensorflow deep-manifold-traversal github.com
Clipped on: 2016-06-11

Skip to content
Tensorflow implementation of Deep Visual Analogy-Making
Python Shell
Latest commit 39e72d0 on Feb 14 Image (Asset 2/8) alt= update loss diagram


Deep Visual Analogy-Making

Tensorflow implementation of Deep Visual Analogy-Making. The matlab code of the paper can be found here.

Image (Asset 3/8) alt=

This implementation contains a deep network trained end-to-end to perform visual analogy making with

  1. Fully connected encoder & decoder networks
  2. Analogy transformations by vector addition and deep networks (vector multiplication is not implemented)
  3. Regularizer for manifold traversal transformations

This implementation conatins:

  1. Analogy transformations of  shape  dataset
    • with objective for vector-addition-based analogies (L_add)
    • with objective for multiple fully connected layers (L_deep)
    • with manifold traversal transformations



First, you need to download the dataset with:

normal$ ./download.sh

To train a model with  shape  dataset:

normal$ python main.py --dataset shape --is_train True

To test a model with  shape  dataset:

normal$ python main.py --dataset shape 


Result of analogy transformations of  shape  dataset with fully connected layers (L_deep) after 1 day of training.

From top to bottom for each : Reference, output, query, target, prediction, manifold prediction after 2 steps, and manifold prediction after 3 steps.

  • Change on angle

Image (Asset 4/8) alt=

  • Change on scale

Image (Asset 5/8) alt=

  • Change on x position

Image (Asset 6/8) alt=

  • Change on y position

Image (Asset 7/8) alt=

(in progress)

Training details

Image (Asset 8/8) alt=



Taehoon Kim / @carpedm20