Watch the pitch videos from last year's five finalists!
These start-ups are bringing new innovations to the computer vision and visual AI market.
We were pleased to recognize and congratulate Retrocausal and Opteran for their achievements at the 2021 Embedded Vision Summit. During the final round of competition on May 27, 2021, it was clear they stood out for their technical innovation, as well as excellence in their business plans, teams and market potential. Take a look below at the brilliant ways our winners and other finalists from 2021 have used computer vision and visual AI to solve real-world problems.
Retrocausal - Judges' Award Winner!
Retrocausal is the industry leader in systems that help manufacturing workers avoid assembly mistakes, be more efficient at their daily jobs, and improve the processes they drive.
Retrocausal is the industry leader in intelligence augmentation systems that help manufacturing workers avoid assembly mistakes, be more efficient at their daily jobs, and improve the processes they drive. Their RetroActivity “Virtual Supervisor” platform uses an AI-enabled camera to analyze the tasks performed on your assembly line in real-time, offering audible and visual alerts when mistakes happen to ensure the operator completes the build correctly before the next step in the process. Their platform also continuously performs time studies on assembly processes to help your business accelerate Kaizen events.
Opteran - Audience Choice Award Winner!
Opteran is a brain biomimicry spin-out from the University of Sheffield, leveraging over eight years of research and 600 million years of evolution to understand how insect brains navigate and enable a new dawn for autonomy in machines.
Opteran is a brain biomimicry spin-out from the University of Sheffield, leveraging over eight years of research and 600 million years of evolution to understand how insect brains navigate and enable a new dawn for autonomy in machines. By capturing nature’s biological systems onto silicon, they enable machines to see, sense, navigate and make decisions better than humans can code. They call this ‘natural intelligence’. It breaks through the limitations of today’s deep learning AI to empower smaller, lighter, ultra-low power solutions that are an order of magnitude more efficient and operate on the edge of the network without requiring a data center or extensive training—just as in nature.
Deep Netts
Deep Netts is a deep learning development platform that makes it easy for non-experts to create and integrate machine vision in Java into existing Java-based applications.
Deep Netts is a deep learning development platform that makes it easy for non-experts to create and integrate machine vision in Java into existing Java-based applications. It provides an integrated development environment for training deep learning models for visual recognition tasks and a pure Java library that enables embedding AI solutions in other products. The community edition of the library is open-source and was chosen as a reference implementation for the official Java Standard for Visual Recognition API, JSR381. The innovative nature of this product focuses on the simplification and integration of the deep learning/machine vision development workflow for Java application developers. The platform provides an agile development tool that facilitates large-scale deployment and wide adoption of ML applications.
Monument
Monument delivers cloud-level photo organization using its hardware optimized AI software to provide a clutter-free photo browsing experience.
Traditional AI models are compute-intensive and designed to run on cloud-based architectures. Monument delivers cloud-level photo organization using its hardware optimized AI software to provide a clutter-free photo browsing experience. Today, Monument is serving over 20,000 active users and creates a new layer in photo management with a focus on privacy.
- Efficient: AI models are hardware optimized, running on NPU for increased performance.
- Cost effective: Local inference eliminates the need of sending data and processing them in the cloud.
- On-device processing: With the on-device AI, user content is processed locally. This provides a better user interface with uncompromised security.
Nomitri
Nomitri focuses on bringing deep learning-enhanced video analysis to embedded devices.
Nomitri focuses on bringing deep learning-enhanced video analysis to embedded devices. They optimize a diverse set of deep learning models serving different computer vision tasks for embedded deployment. Their main software stack wraps these models in highly optimized C++ libraries and combines them with computer vision algorithms in a modular way and can be deployed easily to both mobile and embedded platforms. Nomitri’s vision is to deliver fully embedded, privacy-preserving and lightweight visual AI solutions that enable a host of potential customers to profit from AI capabilities. Their applications focus currently on the stationary retail space.
