oc-bench

Simple Codebases for Benchmarking Object-Centric Architectures

Installation

pip install -r requirements.txt

How to Run

python scripts/train_steve.py
# example
# `python scripts/train_steve.py --level A --image_size 128`
# check the arguments for experiments with '-h' option

Citations

[1] Singh, G., Wu, Y. F., & Ahn, S. (2022). Simple unsupervised object-centric learning for complex and naturalistic videos.

[2] Kipf, T., Elsayed, G. F., Mahendran, A., Stone, A., Sabour, S., Heigold, G., ... & Greff, K. (2021). Conditional object-centric learning from video.

[3] Seitzer, M., Horn, M., Zadaianchuk, A., Zietlow, D., Xiao, T., Simon-Gabriel, C. J., ... & Locatello, F. (2022). Bridging the gap to real-world object-centric learning.

Reference Codebases

[1] singhgautam/steve

[2] a-imamshah/savi-pytorch