Skip to content

drv-agwl/phyre-1

This branch is 21 commits ahead of, 2 commits behind augustinharter/phyre:master.

Folders and files

NameName
Last commit message
Last commit date

Latest commit

9959f05 · Nov 30, 2020
Nov 11, 2020
Nov 5, 2020
Sep 11, 2020
Nov 5, 2020
Nov 19, 2020
Nov 14, 2020
Nov 11, 2020
Nov 5, 2020
Nov 5, 2020
Nov 11, 2020
Nov 19, 2020
Jul 24, 2020
Oct 5, 2020
Jul 24, 2020
Jun 19, 2020
Sep 17, 2020
Nov 5, 2020
May 9, 2020
Sep 10, 2020
Jul 24, 2020
Nov 5, 2020
Nov 5, 2020
Nov 5, 2020
Nov 11, 2020
Nov 30, 2020
Aug 31, 2020
Jul 24, 2020
May 1, 2020
Sep 17, 2020
Nov 11, 2020
Nov 30, 2020
Nov 30, 2020
Aug 30, 2020
May 9, 2020
Nov 5, 2020
Aug 14, 2020
Jun 19, 2020
Nov 30, 2020
Aug 31, 2020
May 11, 2020
Sep 9, 2020
Nov 11, 2020
Nov 5, 2020
Nov 5, 2020
Nov 5, 2020

Repository files navigation

PHYRE

Project space for tackling the PHYsical REasoning benchmark test (phyre.ai)

PHYRE roullouts

use sampler.py to collect solving phyre rollouts from the original simulator

Pymunk Simulator

For concept exploration and proofing pymunk is used to imitate the pyhre simulator. Pymunk can be completely intercepted and customized to try out concepts and to gather custom data.

to generate raw rollouts with full scenes customize and run:
python action_rollouter.py [num_of_rollouts]

to extract training data from raw scenes run:
python scene_extractor.py

Imaginet: Spatial CNN for solving path prediction

SCNN generates feasability map of where the goal ball can be found in a solving path, based on the inital scene (before any action was taken). This is further passed into another modul which is trained to predict the action path which could further be used to extract the actual action position (work in progress)
From left to right: True Target Trajectory, Generated Target Trajectory, Base Trajectory without Action, True Action Trajectory, Generated Action Trajectory Results

ProposalNet: NN to predict paths of dynamic objects

Results from SfM2 Learner

Results

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 98.9%
  • Shell 1.1%