Edgetensor has developed an edge AI-based solution which can tell whether the driver is tired, drowsy or is distracted by his smartphone
Distracted driving is the prime reason for fatalities on the road, necessiating a Driver Monitoring Solution (DMS) that can alert the driver on time and prevent fatal accidents. Edgetensor, a Texas-based startup with a development centre in Bengaluru, has developed a unique Artificial Intelligence (AI) -based driver monitoring system for self-driving cars.
The DMS is designed to run robust AI processing software on low-cost, commodity hardware using any camera. It uses an AI inference engine optimised for low power devices, which delivers a high-accuracy cost-effective way to monitor passengers in real-time. The system offers face tracking, head posture detection, eye and mouth tracking, and gaze and iris tracking, to make sure a person is alert. It can even identify the mood of the driver based on facial expressions.
Edgetensor’s DMS for fleet and automotive markets classifies whether the driver is tired, drowsy or distracted by his smartphone. Alerts are generated by vibrating the driver’s seat or steering wheel or sending audio alerts that can save lives and prevent fatal accidents. “Our short-term and mid-term focus is on non-automotive markets and long-term is on automotive markets,” said Rajesh Narasimha, co-founder and CEO, Edgetensor.
Edgetensor has used edge-computing as a solution. In simple terms, an edge device is one that can perform computations locally without sending data to the cloud for processing and bypasses cloud computing roadblocks such as speed, latency, privacy, reliability and safety concerns. Edgetensor’s solution works with a wide range of edge hardware devices and is compatible with most of the cameras.
“Edge computing is growing rapidly and with many large companies investing in low-power edge compute devices, there is a need for software innovation to run AI algorithms that are compute, memory and data intensive on off-the-shelf commodity devices in a scalable and affordable way,” added Narasimha.
The software development kit (SDK) handles a number of functions, including face tracking with occlusion reasoning, head-pose tracking and location, gaze direction and location, age and gender, face recognition, face identification or verification and emotion recognition. And beyond facial tracking, it can also be used for human detection and pose tracking, vehicle detection and identification and license plate recognition. This means the tech has applications in markets such as video security as well as applications in mobility, personalised robotics, smart cities, smart retail and marketing.
“An important aspect of edge computing is the volume of data sent to the cloud, more so in case of video data from camera sensors, and the power consumed to process the data is high which is reflected in the cost of cloud services,” said Soumitry J Ray, co-founder and CTO, Edgetensor. “Edge devices being inherently low-powered alleviate these issues by crunching the data themselves and thus reduce the cloud footprint of AI algorithms.”
Narasimha and Ray obtained their PhDs from the Georgia Institute of Technology and between them, share more than 75 patents and publications and over a decade of industry experience. Both were part of metaio, a German AR startup acquired by Apple in 2015. The edge computing market is expected to touch $34 billion by 2023 with the Indian market estimated at around $1.5 billion.
“Our two main use cases are distracted driver monitoring, in-cabin monitoring and in-car digital advertising,’’ added Narasimha. The company raised funds from SRI Capital and Aristos Ventures recently and is currently focusing on the US, EU, Middle East, Singapore and Japan. “Our business model is IP licensing, value addition and subscription,” he added.