LatticeFlow raises $12 million to eliminate computer vision blind spots • TechCrunch

LatticeFlow, a startup that spun off Zurich’s ETH in 2020, helps machine learning teams improve their AI vision models by automatically diagnosing problems and improving both the data and themselves Models. The company today announced that it has raised a $12 million Series A funding round led by Atlantic Bridge and OpenOcean, with participation from FPV Ventures. Existing investors btov Partners and Global Founders Capital, which led the company’s $2.8 million seed round last year, also participated in the round.

As LatticeFlow co-founder and CEO Petar Tsankov told me, the company now has more than 10 customers in both Europe and the US, including some big businesses like Siemens and organizations like Sugar Iron Federation Switzerland, and currently operates pilots with quite a bit more. It was this customer demand that drove LatticeFlow up at this point.

“I was in the United States and I met some investors in Palo Alto, Tsankov explained. “They saw the bottleneck we had with new customers coming in. We actually have machine learning engineers supporting our customers and that’s not how you should run a company. And they said, ‘OK, take 12 million dollars, bring these people in and expand.’ It was definitely a great time because when we talked to other investors we saw that the market had changed. “

As Tsankov and co-founder CTO Pavol Bielik have noted, most businesses today have a hard time getting their models into production and then, when they do, they often realize they don’t. works as well as they expected. LatticeFlow’s promise is that it can automatically diagnose data and models to find potential blind spots. In partnership with a large medical company, for example, their model and dataset analysis tools quickly found more than half a dozen critical blind spots in their modern manufacturing models.

The team notes that it is not enough to just look at the training data and ensure that there is a diverse set of images – in the case of the visual models that LatticeFlow specializes in – but also to test the models. Figure.

LatticeFlow Founders Team

LatticeFlow founding team (from left to right): Prof. Andreas Krause (scientific advisor), Dr. Petar Tsankov (CEO), Dr. Pavol Bielik (CTO) and Prof. Martin Vechev (scientific advisor). Image credits: LatticeFlow

If friend only look in the data – and this To be one basic discriminator Because OFFERatticeFlow because we do not only Find Standard data problem alike Labeling problem or poor-quality sample, but also model blind the point, which to be the situations where the model to be Tsankov explained. “Once the model To be Ready, we maybe Take it it, find much data model problem and Help companies repair it.”

For example, models will often find hidden correlations that can confuse the model and skew the results, he noted. For example, when working with an insurance client who used an ML model to automatically detect dents, scratches, and other damage in an image of a car, the model often labeled an image. The photo has a finger on it, it’s a scratch. Why? Because in the training set, the client often takes a close-up of a scratch and points it with a finger. Unsurprisingly, the model would then correlate the “finger” with the “scratch”, even if there were no scratches on the vehicle. The LatticeFlow team argues that those are issues beyond generating better labels and needing a service that can look at both the model and the training data.

LatticeFlow detects deviations in data to train AI models that check for car damage. Because people often point to scratches, this causes models to learn that fingers indicate damage (a false feature). This issue is fixed by a custom upscaling method that removes fingers from all images. Image credits: LatticeFlow

It is worth noting that LatticeFlow itself is not in the training domain. The service works with pre-trained models. For now, it is also focused on offering its service as an on-premises tool, although it could also offer a fully managed service in the future, as it uses fresh capital to hire aggressively, both to better serve its existing customers and to build its product portfolio.

Sunir Kapoor, operations partner at Atlantic Bridge, said: “The painful truth is that today, most large-scale AI modelling implementations do not work reliably in the real world. “This is largely due to the lack of tools to help engineers effectively address critical AI data and model errors. However, this is also why the Atlantic Bridge team came to the decision to invest in LatticeFlow so explicitly. We believe the company is poised for massive growth, as it is currently the only company that automatically diagnoses and repairs AI data and model errors at scale. “

Source by [author_name]


News7h: Update the world's latest breaking news online of the day, breaking news, politics, society today, international mainstream news .Updated news 24/7: Entertainment, the World everyday world. Hot news, images, video clips that are updated quickly and reliably

Related Articles

Back to top button