Some of the New Products on Show
DAY 2: Neurocat spotlights high-fidelity data augmentation
Neurocat
Neurocat, a Berlin-based startup, is displaying its Augmentation as a Service, which is tailored to meet the needs of perception component developers in the autonomous vehicle industry by providing customized data augmentation that adapts rapidly to diverse environmental conditions and regulatory requirements.
Madhusudhan Vasavakkala, go-to-market lead at Neurocat, praised the exhibition: “The ADAS & AV Expo is an unparalleled opportunity to engage with the ADAS and autonomous driving community in Europe. This event brings together key stakeholders, fostering an environment for meaningful dialogue and exchange. It opens doors to new business opportunities and allows us to stay updated on the latest advancements and innovations in the ADAS and autonomous driving industry, which is crucial for the continuous evolution of our solution.”
Data augmentation is essential for enhancing original data sets to reflect a wide range of conditions within the operational design domain (ODD). This process is crucial for the efficient and effective training of machine learning models. Neurocat says its approach addresses the complexity of making data visually realistic, statistically representative and scalable.
At the core of Neurocat's service is the Aidkit software, which combines scalability with customization. The service process includes a discovery phase to understand client needs, a design phase using Aidkit to create the augmentations and a delivery phase to ensure quality and suitability.
“Our approach leverages established image quality metrics to ensure that our augmentations produce activation patterns in machine learning algorithms similar to those created by real-world data,” added Vasavakkala. “Furthermore, our solution integrates seamlessly into perception developers' workflows, enabling direct access to scalable and customizable augmented data for use cases ranging from algorithm training to safety validation.”
Clients leveraging Neurocat’s customized, context-aware augmented data have reportedly seen significant improvements in their perception algorithms. Specifically, they have observed a reduction in false positives and achieved enhanced performance in adverse weather conditions by using augmented data in retraining exercises.
“Leveraging Neurocat’s high-fidelity data augmentation, ADAS and AD perception teams can streamline data collection campaigns, saving on budgets and resources while accelerating AD development cycles,” said Neurocat CEO Florens Gressner.
Visitors can learn more about Neurocat's high-quality, customized data augmentation solutions at Booth 6724.