Berlin-based Hasty, a provider of next-gen annotation tools for vision AI practitioners, has raised $3.7M (approx €3.1M) in its Seed round of funding led by Shasta Ventures, alongside Coparion and iRobot Ventures.
The company will use the raised capital to accelerate product development and expand its customer base across Europe and North America.
About Hasty
Founded in 2018 by Tristan Rouillard, Alexander Wennman, and Kostya Proskudin, Hasty is a Vision AI company that helps humans teach machines to see the world.
The startup supports Vision AI practitioners by developing annotation tools that are supported by a community of machine learning engineers, data scientists, and software developers.
Hasty’s next-generation annotation tool labels data in one-tenth of the time through self-learning AI-assistants and agile machine learning which provides rapid feedback so engineers can validate and adapt models as they work.
How Hasty automates to train AI models and improve annotation
Computer vision (CV), or vision AI, helps computers to gain a high-level understanding of digital images or videos. It also has the potential to change how daily tasks are performed – from diagnosing diseases to tracking packages through the supply chain, to analysing advanced materials worldwide.
Currently, approaches to data labelling are too slow, claims the company. And that they don’t train and update the model while labelling, which means 80% of a data scientist’s time is spent finding, cleaning, and organising the “ground truth” data they use to train their neural networks. Therefore, more than half of the vision-based AI projects never make it into production.
Machine learning engineers often have to wait for 3-6 months for the first results to see if their annotation strategy and approach is working because of the delay between labelling and model training.
This is where Hasty’s next-gen tool speeds up the process because it trains the model as the label. The more the tool is used, the faster the labelling becomes.
Rapid feedback
According to Tristan Rouillard, co-founder, and CEO of Hasty, “There are over 750,000 machine learning practitioners today who are working on vision AI topics and spending the majority of their time managing data rather than building and tuning neural networks. That is why we have created a next-gen annotation tool and community that reduces data prep and management time by 70%.”
Hasty’s agile AI delivers rapid feedback so engineers can adapt and validate their approach as they work.
Hasty’s highlighted features and capabilities include:
- Hasty AI Assistants: it predicts labels just after a few annotated images. Its neural networks learn while engineers build their datasets enabling engineers to go from idea to first model results in a single day
- Hasty Error Finder & Manual Review: this detects hidden annotation errors from misclassifications, artefacts, missed objects, and poor segmentation, helps users to fix it within minutes instead of spending days looking for errors
- Hasty Rapid Feedback Loop: this gives short feedback cycles, enabling users to iterate quickly to validate and adapt their approach as they label. Now users can adapt their annotation strategy, test different approaches, and be confident they are creating the right data in weeks rather than months.
Image credit: Hasty
01
How Gen Z affects B2B social media