Home / News / Artificial Intelligence / V7 spends $33M to automate computer vision AI model training data

V7 spends $33M to automate computer vision AI model training data

Ironically, creating such AI faces a significant scalability issue. Artificial intelligence (AI) promises to assist, and perhaps even replace, humans in carrying out routine jobs and solving challenges that humans have been unable to address. It has to source and absorb ever-larger data troves because it is only as good as the models and data used to train it. However, the time and money required to annotate and manipulate that training data slows down the task or, maybe, both.

A company by the name of V7 Labs thinks it has made progress in this area. It has successfully created training models for automating the development of those models. After noticing a high demand for its services, it is today announcing $33 million in funding to support its growth.

Today, V7 is concentrating on computer vision and aiding in object identification. It asserts that with just 100 manually annotated examples, it can learn what to perform.

It has gained significant popularity in the medical and scientific industries, where its platform is being used to aid in the training of AI models to expedite processes such as the detection of tumors and other abnormalities on scans. V7 is also beginning to see activity from tech and tech-savvy businesses looking at ways to deploy its technology in a range of other applications, such as businesses developing tools that can generate images from commands given in natural language and business applications. GE Healthcare, Paige AI, Siemens, as well as other Fortune 500 organizations and larger privately-held businesses are among the over 300 clients on the list, though the company isn’t releasing a complete list of those using and assessing its technology.

Air Street Capital, Amadeus Capital Partners, and Partech—three prior backers—as well as a number of well-known figures in the machine learning and AI industries are also participating in this round, which is being co-led by Temasek and Radical Ventures. They consist of Ashish Vaswani, a co-founder of Adept AI who previously worked at Google Brain where he created Transformers, Francois Chollet, an inventor of the open-source Keras Python neural network library, Oriol Vinyals, a principal research scientist at DeepMind, Jose Valim, an inventor of the Elixir programming language, and unnamed others from OpenAI, Twitter, and Amazon.

According to CEO Alberto Rizzoli, this is the largest Series A fundraising round in this category to date. It will be used to expand the company’s commercial operations and hire more engineers to handle a surge in consumer interest, with a focus on the United States. He declined to comment on the startup’s valuation, but it has now received about $36 million, and my understanding is that it is currently valued at about $200 million.

Additionally, Rizzoli steered clear of discussing income data but noted that ARR increased threefold in 2022.

In order to address the larger field of AI modeling and to increase the effectiveness of training AI data, a number of other companies have appeared. One of V7’s more direct rivals is SuperAnnotate, which, according to PitchBook, has garnered over $18 million. (V7 even provides a comparison of the two services.) Others include Labelbox, which collaborates with businesses like Google and others on AI labeling, Scale AI, which initially focused on the automotive industry but has since branched out into a number of other areas, and Hive, which is currently valued at about $2 billion.

V7 and the others are relying on the fact that the training model is ineffective and can be improved. V7 is named in reference to AI being the “seventh” area for processing images after the six areas in the human brain that make up its visual cortex (V1 through V6).

Automation is a unique selling point of V7. It has created a model to automate the process of labeling, recognizing when something is incorrectly labeled, rethinking categorizations, and other tasks that take up roughly 80% of an engineering team’s time.

The method it has developed is known as “programmatic labelling”; according to Rizzoli, who co-founded the company with its CTO Simon Edwardsson, it only requires 100 “human-guided” examples for its automated labeling to begin working. General-purpose AI and its own algorithms are used to segment and label images.

Investors wager that speeding up the process of developing and using AI models will increase the company’s revenue. “Computer vision is being used widely throughout industries, bringing innovation and technological advancements, and a rapidly expanding $50 billion business. In a statement, partner at Amadeus Capital Partners Pierre Socha stated, “Our thesis for V7 is that the breadth of applications and the pace at which new products are expected to be introduced in the market call for a centralised platform that connects AI models, code, and humans in a looped ecosystem.

Although V7 refers to the process as “autopilot,” co-pilot may be a more accurate term because it is intended that anything marked as unclear be sent back to humans for evaluation and review. Instead of replacing those people, it merely makes it simpler for them to do their assignments more quickly. (It occasionally performs better than humans, so using both together could be useful to double check each other’s work.) The picture training on a scan to detect pneumonia is demonstrated below.

Rizzoli noted that the decision to initially focus on the medical industry was made in part to keep the startup’s feet on the ground and to focus on a market that might not have built this kind of technology in-house but would definitely want to use it. This is because there are many areas where AI is being applied to improve how images are processed and used.

He added, “We choose to focus on verticals where a lot of work on visual processing is being done, but by humans, or where AI-based apps are already being commercialized.” We didn’t want to be attached to moonshots or projects with large R&D budgets because those indicate that someone is trying to totally solve the problem on their own, they are working on something more specialized, and they might want to use their own technology rather than that of a third party like us.

In addition to businesses looking for “their own secret sauce,” Rizzoli said that projects may remain in the lab and never leave. Instead, he remarked, “We are working on actual applications.

In another sense, the startup symbolizes a change in how businesses are acquiring and utilizing information. Investors believe that the framework V7 is developing may eventually alter how data is ingested by those businesses.

According to Parasvil Patel, a partner at Radical Ventures, “V7 is well-positioned to become the industry-standard for managing data in modern AI workflows.” With this round, Patel joins the board of V7.

“AI can now solve a huge variety of issues, and this number is expanding quickly. In order to deliver outstanding products that continuously improve and adapt to real-world needs, businesses of all sizes must have best-in-class data and model infrastructure, said Nathan Benaich of Air Street Capital in a statement. “In this situation, V7’s AI Data Engine excels. Customers rely on V7 to deliver reliable AI-first products faster than ever before, regardless of the industry or application. V7 incorporates multiplayer workflows from data to model to product that reflect the rapidly evolving best practices in the sector.

About Chambers

Check Also

The BMW Vision Neue Klasse X has circular battery cells, four advanced computer systems, and a three-dimensional grille

BMW has just unveiled the Vision Neue Klasse X, a novel concept car that the …