If you want to set up an early warning system for detecting forest fires, there’s a type of multimodal sensor which can accurately detect the visual, sonic and olfactory signs of fire: a human being equipped with eyes, ears and nose.
But it’s clearly not viable to station hundreds of humans throughout a forest wilderness, waiting for those telltale signals. What’s needed is human-like sensory intelligence – but without the humans. And that’s what Grenoble, France-based MountAIn has created with its AI end-to-end infrastructure.
MountAIn’s technology is already in use: since July 2024, a successful pilot in the south of France has been monitoring for potential wildfires, using an array of AI-enabled smart cameras to detect smoke and trigger real-time alerts before fires escalate. Just today (11/05/2025) MountAIn was announced as a honoree for the CES Innovation Awards.
Forest fires – an increasing hazard worldwide due to climate heating. (Image credit: StevenTerblanche under Creative Commons license)
The heart of the MountAIn system is IBEX, a ‘digital watchman’ trained to monitor for the specific parameters appropriate to individual use cases – not only forest fire detection, but also person counting, wildlife and livestock monitoring/tracking, intruder detection, flood monitoring, and workplace safety. The IBEX smart camera continuously monitors its surroundings at 10-20 frames per second, tracking activity and sending real-time alerts when pre-defined conditions are met.
The IBEX unit operates without an external power supply, relying instead on solar energy, and connecting to a remote server or the cloud via satellite communication to send short updates. The satellite connectivity is not, however, a conduit for exchanges with a cloud AI server: AI processing – determining whether the camera is seeing smoke blowing in the air, for instance – is performed locally, on the device.
To provide this functionality, MountAIn has performed a difficult feat of engineering – the IBEX system features a combination of:
- Ultra-low power consumption to ensure round-the-clock operation on the energy that can be harvested from the sun and stored in a small battery
- Robust and effective local processing of neural networks which perform image classification
To optimize the power/performance trade-off, MountAIn chose the world’s leading microcontrollers for AI: the Ensemble family from Alif. The IBEX camera unit is based on an Ensemble E3, an MCU which has dual-core combinations of a high-performance (400MHz) Arm® Cortex®-M55 paired with an Arm Ethos™-U55 neural processing unit (NPU), and a high-efficiency 160MHz Cortex-M55, associated with a second Ethos-U55 NPU.
The value of the E3 in this application is due to the unique Alif combination of high-performance and high-efficiency AI/control blocks: the IBEX system uses the high-efficiency block in always-on mode to continuously monitor its environment for potentially relevant signals; it switches on the high-performance block to run a high-accuracy image classification network only on relevant data samples.
This has allowed MountAIn to achieve astonishingly low power consumption in the IBEX system, enabling it to be deployed practically indefinitely in remote locations where no wired power supply exists, and where it would be difficult and expensive to replace a discharged battery.
The Ensemble device is absolutely central to MountAIn’s success. As the company itself says, ‘Alif is the only supplier that provided a device powerful and low-power enough to do what was required in the IBEX device.’
This example shows that effective AI at the edge is possible today – with the right choice of hardware.