ADEPT-Dataset-Release

This is the dataset generation code for ADEPT (Approximate Derenderer, Extended Physics, and Tracking).

Modeling Expectation Violation in Intuitive Physics with Coarse Probabilistic Object Representations

Kevin Smith*, Lingjie Mei*, Shunyu Yao, Jiajun Wu, Elizabeth S. Spelke, Joshua B. Tenenbaum, Tomer Ullman (* indicates equal contribution)

Preprint

Paper BibTeX Website

For the model, see ADEPT-Model-Release

Prerequisites

• Linux
• Python3
• Blender as a python module
• Other modules required specified in requirements.txt

Getting started

1. Clone this directory

   git clone https://github.com/JerryLingjieMei/ADEPT-Dataset-Release
   cd ADEPT-Dataset-Release

   And replace CONTENT_FOLDER in utils.constants and phys_sim/data/builder/collect_obj.sh with the absolute path to your directory.

2. Create a conda environment for ADEPT Dataset Generation, and install the requirements.

   conda create --n adept-dataset
   conda activate adept-dataset
   pip install -r requirements.txt

For installation of Blender as a python module, see Blender wiki.

You may also try using Blender's bundled python, by replacing `python *.py --arg1 --value1` with `blender -b --python -- --arg1 --value1`

1. (Optional) If you have multiple machines, you may change get_host_id in utils/misc.py to reflect the id of your machine. With that in hand, you may speed up all following processes by using --stride N arguments, where you have N machines with consecutive ids.

2. To render ShapeNet objects, please download ShapeNet Core V2 from its official website. Change SHAPE_NET_FOLDER in phys_sim/data/builder/collect_obj.sh to the path of ShapeNet meshes, and run thar script.

   To turn them into .blend files, run

   # Single machine
   python3 render/data/builder/collect_blend.py #Map phase
   python3 render/data/builder/collect_blend.py --reduce #Reduce phase
   # Multiple (e.g. 8) machines
   python3 render/data/builder/collect_blend.py --stride 8 #On each machine
   python3 render/data/builder/collect_blend.py --reduce --stride 8 #On a single machine

Dataset generation

1. Generate training set (e.g. with 1000 videos) by running

   # Single machine
   python3 dataset/generate_train.py --end 1000
   # Multiple (e.g. 8) machines
   python3 dataset/generate_train.py --end 1000 --stride 8 #On each machine

2. Generate human test set by running

   # Single machine
   python3 dataset/human/generate_human.py --end 1000
   # Multiple (e.g. 8) machines
   python3 dataset/human/generate_human.py --end 1000 --stride 8 #On each machine

Evaluation

1. Evaluating the relative accuracy on human test set. If you have a experiment output folder that contains .txt files containing the scores of all human test cases, run

   python3 dataset/human/collect_reuslts.py --summary_folder ${SUMMARY_FOLDER} #Score in SUMMARY_FOLDER/results
   python3 dataset/human/collect_reuslts.py --summary_folder ${SUMMARY_FOLDER} --output_folder ${OUTPUT_FOLDER} #Custom output folder

   Or you may get the relative accuracy from a json file that contains a dictionary mapping case name to its score:

   python3 dataset/human/collect_results.py --summary_file ${SUMMARY_FILE} --output_folder ${OUTPUT_FOLDER} #Custom output folder