Akridata announces the availability of a new AI platform for visual data
LOS ALTOS, Calif., September 8, 2022 – Akridata, an AI platform for visual data, announced the first availability of its platform that provides data science teams with the tools to easily explore, search, analyze and compare data visuals to improve datasets and improve model training.
The volume of visual data has grown at an unprecedented rate, with more and more cameras being continuously deployed around the world. However, the tools and procedures to process this massive amount of data have not kept pace with the growth. For many, data ingestion and curation remains a largely manual and time-consuming project.
“Most academic datasets fall far short of industry needs,” said Helge Jacobsen, senior deep learning engineer at Veo. “In our case, we have rather small objects in rather large images. It’s probably a common problem, but in academia it’s kind of niche.
Akridata Data Explorer is the first platform designed to focus solely on processing visual data in the ML lifecycle. Founded by a team of serial entrepreneurs with deep technical expertise in solving image processing problems, Akridata quickly realized while working with computer vision data science teams that the biggest challenges lay in the search, grouping and selection of visual data to accelerate model accuracy.
“The demand for data scientists and the tools they need will only increase over time,” says Vijay Karamcheti, CEO and co-founder of Akridata. “There are 50 billion cameras in the world, and the data they produce is growing exponentially, so it’s imperative that data scientists have the tools to navigate and analyze this data effectively.”
The initial release of Akridata Data Explorer provides data scientists with tools throughout the MLops lifecycle:
- Tools to visualize and explore large datasets as embedding-based clusters (e.g. spotting different actions by vehicles, people, etc.).
- Data search tools (eg, search for additional instances of found objects within a user-specified bounding box).
- Identification of novelty datasets to reduce class imbalances and data labeling expense.
“Having worked with data scientists to create models for computer vision applications, we understand the challenges they face,” said Sanjay Pichaiah, VP of Product and GTM at Akridata. “For AI models to be production-grade, choosing the training datasets is just as important, if not more so, than the model parameters. There is an urgent need for tools to help scientists in data to make intelligent and informed data selection.”
Akridata was launched in 2018 by a team of serial entrepreneurs with a vision to solve a unique set of visual data problems in the AI training lifecycle. The team collectively brings years of experience ranging from deep engineering, basic research and business experience and has a track record in building and scaling startups.