Infinity AI raises $5M to build synthetic training data
Infinity AI, a startup that generates automated synthetic training data, has announced a US$5mn seed round led by Matrix with participation from founders and operators from companies like Snorkel AI, Tesla, and Google.
AI models are only as good as the data they are trained on, but traditional data collection and annotation processes - fueled by human labour - are notoriously expensive and slow. Research suggests enterprise data scientists spend over 80% of their time gathering, organising, and labelling the training data they use, and bottlenecks in training data are a primary reason that computer vision projects get cancelled.
Synthetic data - or data that is generated via simulation rather than collected by a sensor - leapfrogs the data collection and annotation steps. Gartner recently placed synthetic data at the top of its list of strategic predictions for 2022 and beyond, saying that "by 2024, 60% of the data used for the development of AI and analytics projects will be synthetically generated."
Generative AI-powered tool bring capabilities to the next level
Infinity's platform allows engineers to upload a single real-world video and turn that into hundreds of similar and perfectly labelled synthetic videos. Infinity uses a combination of physics-based simulations and generative techniques to do this. For example, its self-serve API allows engineers to generate hundreds of videos that meet desired statistical distributions of camera location, lighting conditions, avatar appearances, and more.
Infinity says its generative tools, currently in beta, bring those capabilities to the next level. They include a Stable Diffusion-based inpainting tool that massively augments scenes and another generative tool that adds infinite clothing textures to avatars. Infinity will release these tools publicly in 2023.
“Using Infinity AI, we can get new products out the door faster. And, our ML engineers are happier because they get to spend more time on the fun part of model development,” says Harishma Dayanidhi, Co-Founder and VP of Engineering at Voxel, one of Infinity AI's early customers.
“We want to make it easy for ML teams to start working with synthetic data,” says Lina Avancini Colucci, one of the founders of Infinity AI. “The ML community has a scarcity mentality with regards to data today. Synthetic data turns this into an abundance mentality. Infinity is democratising access to training data since this is the biggest roadblock to progress in ML today.”