New Imagimob Applications Show the Potential of TinyML

Imagimob recently announced it is using its TinyML platform technology combined with Texas Instruments’ mmWave radar sensors for fall detection and gesture recognition. Users can download Imagimob AI for free, with both fall detection and gesture recognition included in the platform as starter projects. In addition to the fall detection and gesture recognition applications, Imagimob customers can also use the platform to determine audio levels in places like factories or mines, to detect human motion and track workers and elderly people, and for predictive maintenance of many machines and devices. The announcement of TinyML for these two applications shows the vast potential for Machine Learning (ML) technologies to perform on-device analytics of sensor data and perform automated tasks, even when connections are down.

Fall recognition provides great value for monitoring people in numerous scenarios. The sensors can be installed on applications throughout people’s homes, in nursing homes, or on different machines in a factory or at a job site. With TinyML technology, the sensors automatically send an alert to a device, whether that be a phone or a computer, when a fall has been detected so that action can be taken. The automated alert allows for an immediate response to situations that could be dangerous for workers or elderly people who may not be able to get up after falling.

The gesture recognition application is able to recognize six different predefined gestures, although the six gestures have yet to be published. Traditionally, interfaces require the user to touch the screen, which can be problematic while they are driving a vehicle or operating dangerous equipment. Not only does this require the user to divert their attention, but the screen also wears down after so many touches. The hand gesture recognition allows for the user to complete the input they want to provide to the interface without diverting much of their attention. For example, when someone is driving their vehicle and wants to adjust the air conditioning in more modern vehicles, they have to perform multiple touches of the screen to press the small buttons on the screen. With hand gesture recognition, the driver can perform a hand gesture like pointing their thumb downward to indicate they want the temperature to be decreased. “Hands-free” is a term that is widely used to indicate safer usage of phones and devices in vehicles and gesture recognition provides even more hands-free opportunities.

The TinyML market is still in the early stages of adoption, but these two applications show the potential for adoption across many industries. Normal ML applications are performed over the cloud, which takes up a lot of memory and cannot run on embedded hardware. With TinyML, ML can be done on the device and no data ever have to be transferred. More importantly, the recognition process of a problem or an alert is completed faster, allowing an automated action to be taken within milliseconds. Additionally, when the data are processed on the device, actions can be completed even when the network connection is down.

ABI Research recommends that companies implement TinyML technologies in their device design process and budget. ABI Research forecasts that the number of TinyML devices will grow from 15.2 million shipments in 2020 to more than a billion by 2030. The challenge for device Original Equipment Manufacturers (OEMs) attempting to implement ML in their devices is completing this process with limited available power. There is also a heavy push for device miniaturization in the Internet of Things (IoT) space, so room within the device is also a significant constraint. However, ABI Research expects to see significant growth in the adoption of TinyML technology in applications across many industries, especially for condition monitoring and retail applications. The substantial benefits that this technology can provide companies in the form of condition monitoring, predictive maintenance, and people/worker safety, to name a few, will drive innovators to overcome these challenges and provide numerous solutions across the IoT ecosystem.