Deep Visual Analogy-Making
Authors: Scott E. Reed, Yi Zhang, Yuting Zhang, Honglak Lee
NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In experiments, our model effectively models visual analogies on several datasets: 2D shapes, animated video game sprites, and 3D car models. We evaluated our methods using three datasets. |
| Researcher Affiliation | Academia | Scott Reed Yi Zhang Yuting Zhang Honglak Lee University of Michigan, Ann Arbor, MI 48109, USA {reedscot,yeezhang,yutingzh,honglak}@umich.edu |
| Pseudocode | Yes | Algorithm 1: Manifold traversal by analogy, with transformation function T (Eq. 5). Algorithm 2: Disentangling training update. |
| Open Source Code | No | The paper does not contain an explicit statement about releasing its source code for the described methodology or a direct link to a code repository. |
| Open Datasets | Yes | The second is a set of 2D sprites from the open-source video game project called Liberated Pixel Cup [1], which we chose in order to get controlled variation in a large number of character attributes and animations. The third is a set of 3D car model renderings [11]. [1] Liberated pixel cup. http://lpc.opengameart.org/. Accessed: 2015-05-21. [11] S. Fidler, S. Dickinson, and R. Urtasun. 3d object detection and viewpoint estimation with a deformable 3d cuboid model. In NIPS, 2012. |
| Dataset Splits | Yes | We split the data by characters: 500 training, 72 validation and 100 for testing. |
| Hardware Specification | Yes | We thank NVIDIA for donating a Tesla K40 GPU. |
| Software Dependencies | No | We used Caffe [14] to train our encoder and decoder networks, with a custom Matlab wrapper implementing our analogy sampling and training objectives. No version number for Caffe or Matlab is provided. |
| Experiment Setup | Yes | We trained for 200k steps with mini-batch size 25 (i.e. 25 analogy 4-tuples per mini-batch). We used SGD with momentum 0.9, base learning rate 0.001 and decayed the learning rate by factor 0.1 every 100k steps. We trained the models using SGD with momentum 0.9 and learning rate 0.00001 decayed by factor 0.1 every 100k steps. Training was conducted for 200k steps with mini-batch size 25. |