An AI research lab focused on creating simulation software and training pipelines for multimodal generative models.
Currently focusing on improving spatial understanding and motion in video gen models.
Simverse is a synthetic video & image generator with auto-captioning. Simverse allows users to place objects in scenes, add movement, auto-caption, and prompt it to generate videos according to specific requirements. Enables the creation of high-quality training datasets for a wide range of computer vision and NLP tasks.
- Simverse & SimData3D dataset paper + release date late 08/2024
Distributask is a simple way to run distributed jobs, leveraging low-cost hardware from Vast.ai's marketplace. Get your synthetic data, rendering or data processing jobs done cheap and at scale. Deliver final results to Huggingface with one line of code.
If any of the research topics interest you reach out to eric [at] deepai.org
Find us at HuggingFace / Discord
A DeepAI Research Group