Predictive and generative neural networks for object functionality

Information

• Paper topic: Geometry
• Software type: Code
• Able to run a replicability test: True
• Replicability score: 0
• Software language: Python
• License: MIT
• Build mechanism: Not applicable (python, Matlab..)
• Dependencies: tensorflow / numpy / scipy
• Documentation score {0,1,2}: 0
• Reviewer: David Coeurjolly <david.coeurjolly@liris.cnrs.fr>
• Time spent for the test (build->first run, timeout at 100min): 50min

Source code information

• Code URL: https://github.com/Salingo/PGDNN
• Git commit hash: 72c801c890ab8233dbbe4a9cb4cfee8a69e7d494
• Software Heritage link:

Comments

Missing dependencies : numpy and scipy but easily installable. The code does not provide the pretrained models and requires to train three different networks (15000 epochs (2-3min per 10 epochs on the test machine)). I was not able to check the results and thus evaluate if the results match with the paper ones.

If you want to contribute with another review, please follow these instructions.

Please consider to cut/paste/edit the raw JSON data attached to this paper.