Imagimob AI
A system for real time artificial intelligence on the edge

Imagimob AI is a system for Edge AI applications that enables small devices with intelligence and data processing capabilities, to empower you and your customers with actionable insights—in real-time.

A radical approach to Edge AI

When the Internet of Things and Edge AI come together, the possibilities are tremendously exciting. Imagine a product that not only has the capacity to collect data, make informed decisions, learn, and improve—but to do all of these things independently of network connections. The result can bring added safety, convenience and efficiency to the user experience, and even save lives.

However, for many, the reality of incorporating such technology into a single device is too costly and impractical. Even some of the most promising machine learning models, widely applied in the cloud and digital industries, are still using too much computing power and memory to be able to fit the small microprocessors on the market today. And the alternative of engineering customized systems is often far too expensive to bring to fruition.

At Imagimob, we’ve figured out a way to create AI systems that can run efficiently on small devices while still delivering the performance and accuracy typically only found in bigger systems. Combining inspiration from the machine learning field with our own innovations, each of our algorithms is designed with small MCUs, low CPU usage, and low memory consumption in mind.

The Imagimob effect

Imagimob AI software offers high accuracy in real-time, on low-cost hardware with low power consumption. This lends itself to a world of exciting applications, where the AI model is pre-trained with data to create specific instincts for use in all kinds of intelligent products—from fall detection clothing to predictive maintenance tools. The pre-trained application collects data with its sensors, analyzes it for relevant patterns, then sends a signal only when action is needed. 

 

 

Imagimob AI Advantages

Imagimob AI offers a number of advantages compared to cloud-based AI solutions

Actionable insights in real-time

With zero to low latency between sensing, decision, and action, Imagimob AI allows for real time operations, including making critical decisions where milliseconds matter. 

Total autonomy

Imagimob AI makes devices autonomous, which means that they can think for themselves and don’t need to connect to the internet or central processing hub in order to work properly. 

Dynamic intelligence—that keeps getting better

Imagimob AI learns quickly and easily, using small data sets. This means your device only becomes smarter and more sophisticated over time. 

Low power

Having the intelligence on the device lowers the amount of sent data, because it will only send curated data. Sending less data lowers power consumption, because running the software on the device consumes much less power than transmitting all sensor data over Bluetooth LE (BLE) or any other communication method.

Low-cost devices

Imagimob AI Lib can run on low-cost microprocessors, such as ARM Cortex M0. Cost is very important when you want to roll out millions of devices. Imagimob AI Lib requires around 10 kB of RAM on the device.

Low running costs

Imagimob AI operates locally, and only sends data when it matters. This means significantly less power consumption and data transmission costs—crucial for wearable devices. 

Communication-agnostic

Imagimob AI Lib is communication-agnostic, ie it works with all communication technologies in the market, eg Bluetooth LE, 2G/3G/4G/5G and WiFi. Using Imagimob AI Lib together with Low Power Wide Area Networks (LPWAN), such as NB-IoT, Sigfox or LORA is an ideal fit, since they share the same qualities that are important for IoT networks, low power and low cost devices.

Increased privacy

Constantly streaming data to the cloud for storage makes the system vulnerable to privacy violations. By processing data locally on each device, these problems can be avoided entirely. 

Imagimob Motion AI
A software suite for Motion AI on the edge

Imagimob AI Motion uses input from motion sensors such as accelerometers, gyros and magnetometers, and translates this into motions. It includes the following components:

Imagimob AI Capture

Capture comprises a mobile app and a capture device. It’s used to capture synchronised motion sensor data and video during the data capture phase. Once the data capture is ready, it is sent to SensorBeat Analyser for analysis.

Imagimob AI Analyser

Analyser is a software package for analyzing and collecting action from motion sensors and video.  This is where you do the AI learning. You can playback video with synchronised sensor data, in order to identify important patterns. The result of the analysis is the motion pattern database for Imagimob AI Lib.

Imagimob AI Lib

Imagimob AI Lib is a embedded software package that includes the AI engine and a trained AI model. It can be easily integrated into 3rd party hardware platforms. Imagimob AI Lib is fast, accurate and has a very small footprint, which means that it can run on low-cost and low-power hardware devices. For OEM devices the Lib is a source code / pre-compiled library written in ANSI C. The Lib is also available on iOS, Android and Apple watchOS.

Download this fact sheet that explains the actual power consumption for the Lib for a specific application.