Nanoveu Limited (ASX: NVU) has completed comprehensive, industry-standard, benchmark testing for EMASS's ECS-DOT SoC (System on Chip) chipset in the categories of Anomaly Detection and Keyword Spotting.
EMASS’ ECS-Dot chipset has demonstrated exceptional results in the latest benchmarking evaluations covering Anomaly Detection and Keyword Spotting. The ECS-Dot chipset has outperformed leading competitors and peers, including Qualcomm and Himax showcasing significantly superior energy efficiency. These results position the company as an industry front runner in ultra-low-power embedded AI applications.
“We are extremely proud of the performance shown by our ECS-Dot Chipsets in these benchmarks. These results highlight our continuous drive for innovation in ultra-low-power AI solutions. By achieving such energy-efficient performance without compromising on accuracy, we are opening new possibilities for embedded AI applications across various industries,” Mohamed M. Sabry Aly, Founder of EMASS technology, said.
Benchmarking Results and Significance:
Anomaly Detection is a technique used in artificial intelligence and data analysis to identify data points, patterns, or behaviors that deviate significantly from what is considered "normal" or expected. It is particularly valuable in fields where detecting rare or unexpected events is essential for system reliability, security, or personalisation. Anomaly Detection is crucial for applications involving 2D to 3D data conversion based on eye tracking. By detecting abnormalities in user engagement, Anomaly Detection enables systems to adapt dynamically to unexpected behaviours, enhancing accuracy and responsiveness. This technology helps maintain seamless, intuitive, real-time interactions, allowing for a more personalised and immersive user experience.
Keyword Spotting is a fundamental component for interactive language models. This capability enables devices to "wake up" in response to specific speech triggers, activating more complex applications only when needed. This approach allows sophisticated systems to remain idle when not in use, conserving significant amounts of energy. Instead of continuously running background processes, Keyword Spotting allows for an efficient, speech- based AI interface that triggers larger systems on demand. This innovation is especially impactful for devices requiring low-power, “always-on” capabilities, as it ensures energy-efficient performance without sacrificing functionality
In the Anomaly Detection benchmark, Embedded AI Systems achieved the same level of accuracy as the best publicly available results, with an execution time of 1.22 ms but with an energy consumption of only 0.8 microjoules. This represents an over 200x improvement in energy efficiency compared to the best previously reported results by STMicroelectronics.
For Keyword Spotting, the ECS-Dot Chipsets achieved an execution time of 3.9 ms and consumed just 3.07 microjoules of energy per inference, translating to over 10x energy efficiency gains compared to the best results reported by Syntiant.
These significant results underscore the ECS-Dot Chipset’s ability to deliver high performance computation in ultra-low-power settings, making it an ideal choice for real-time, energy-sensitive applications.
Embedded AI Systems has successfully examined all MLPerf Tiny benchmarks and is actively exploring other publicly available benchmarks to showcase the superior efficacy of its ECS-Dot Chipsets across various AI applications. This commitment highlights the company's dedication to incessant improvement and industry leadership in energy-efficient AI solutions and continuous improvement of the EyeFly3DTM platform for delivering glasses-free 3D experiences.
In the first benchmarking results announced3 in October 2024, EMASS demonstrated 20x lower energy consumption compared to peers validating the potential for image conversion in an energy efficient manner which is important for EyeFly3DTM applications.
- In the category of Person Detection EMASS had set a new benchmark for AI processing, achieving a remarkable latency of 5.2 ms per inference while consuming a mere 3.7 microjoules of energy.
- For Image Classification, the company’s solution demonstrated 6.3ms latency per inference while consuming only 5.5 microjoules of energy. This performance is 20x lower in energy consumption compared to previous results reported by other edge AI chipset companies including Synitant, Ambiq, Himax and Maxim integrated.
Embedded AI Systems continues to lead the charge in developing energy-efficient, high-performance AI solutions for IoT, wearable devices, and other real-time applications. The company is actively expanding its portfolio by exploring new benchmarks to further demonstrate the capabilities of the ECS-Dot Chipsets in a wide range of AI driven technologies in areas such as:
- 2D to 3D Conversion: In digital twin technology, it creates real-time 3D models for virtual testing and optimization of industrial systems.
- Self-Navigation of Drones: In Agriculture IoT, drones autonomously monitor crops and manage farm operations.
- Autonomous Driving Systems: In transport and logistics, AI-powered systems allow vehicles to navigate and make decisions independently, optimise transportation and reduce delivery times in urban areas.
- Robotics: In manufacturing, robotics automate assembly lines, enhancing precision and efficiency. In healthcare, robots assist in surgeries and patient care, improving outcomes with minimal human intervention.
- Smart Home Systems: In retail, they enable seamless in-store experiences like voice-activated shopping and personalized services.
- Security and Surveillance: In urban infrastructure, AI-powered surveillance systems monitor public spaces for safety and emergency response.
