hyenal/relate

Official PyTorch implementation of 'RELATE: Physically Plausible Multi-Object SceneSynthesis Using Structured Latent Spaces'.

29
/ 100
Experimental

RELATE helps researchers and engineers create realistic synthetic images and videos of multi-object scenes, adhering to physical plausibility. You provide a dataset of existing images or videos, and it generates new, diverse scenes with objects arranged and interacting in believable ways. This is useful for researchers in computer vision or robotics who need to generate large amounts of varied data for training and evaluation without real-world limitations.

No commits in the last 6 months.

Use this if you need to synthesize large datasets of physically realistic scenes with multiple objects for tasks like computer vision model training or robotics simulation.

Not ideal if you are looking to generate highly artistic or abstract imagery, or if your primary goal is to edit existing images rather than create new ones from scratch.

synthetic-data-generation computer-vision-research robotics-simulation scene-understanding image-synthesis
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 7 / 25
Maturity 16 / 25
Community 6 / 25

How are scores calculated?

Stars

31

Forks

2

Language

Python

License

Category

gan-based-t2i

Last pushed

Nov 06, 2020

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/hyenal/relate"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.