MountAIn's IBEX camera runs computer vision on a microcontroller, powered by the sun, in places with no grid and no network. Roboflow users can now train a model and deploy it there.
“Our customers want vision AI on a hillside or along a remote pipeline, in places with no grid and no network. MountAIn gets the model onto a tiny chip that runs on sunlight,” said Vincent Huard, co-founder and CEO of MountAIn. “We build on Roboflow because no other platform keeps pace with the research the way it does. When a better model comes out, our customers can train it on their own images in Roboflow as soon as it launches and run it on our cameras. Like the rest of the ML community, we rely on Roboflow for access to the best models: RF-DETR, ViTs, ResNets, transformers, the most up-to-date Ultralytics model licenses, and more. A solar-powered camera gets the same state-of-the-art architectures that teams run on GPUs in the cloud. With Roboflow’s open ecosystem and integrations, every MountAIn customer can run their custom models on our devices.”
Roboflow Models, on More Microcontrollers
Most vision models run in the cloud or on Linux cameras, and both get expensive when a deployment grows from one camera to a thousand. On October 21, 2026, MountAIn announced that MountAIn Core now connects Roboflow's community of more than one million developers to low-power microcontrollers, starting with MountAIn's IBEX camera.
Developers can export YOLO11n, YOLOv8n and YOLO26n models trained in Roboflow to IBEX with a few lines of Python, and RF-DETR Nano support is coming soon. MountAIn is focusing first on construction sites, traffic monitoring, critical infrastructure and logistics.
Vision AI Where There Is No Power or Network
Wildfires start in places few people watch: ridgelines, forest interiors, valleys miles from a road. A camera there has no outlet and often no cell signal, so streaming video to the cloud for analysis is off the table.
Grenoble-based MountAIn built its IBEX camera for those sites. IBEX runs on a solar panel and a small battery, processes video on the device, and sends a short alert over satellite when it sees the event it was trained to find. Since July 2024, a pilot in the south of France has used IBEX to spot wildfire smoke and warn crews before fires spread. IBEX was named a CES 2026 Innovation Awards honoree.
Running AI on a Microcontroller
Every milliwatt counts on a solar-powered camera. Device makers have had to choose between microprocessors that run complex vision models but draw more power, and microcontrollers that use little power but leave little room for AI. MountAIn Core, the company's compiler, middleware and runtime, lets the microcontroller do the job: it cuts memory use by 3x and keeps power under 100 milliwatts on Alif Semiconductor hardware.
"Embedded AI shouldn't require a PhD in firmware optimization," Huard said. "Our job is to make the model fit the chip, so teams can focus on what their device needs to see."
Fire detection is IBEX's first use case. Operators can also train it to count people, track wildlife and livestock, detect intruders or watch flood levels. MountAIn also offers the same stack to device makers, through MountAIn Ready and MountAIn Onboard, for products from smart cameras to smart glasses.
Labeling,Training, Deploying in One Place with Roboflow
Each IBEX deployment watches for something different, in scenes no public dataset covers, so each needs a model trained on its own images. MountAIn chose Roboflow so its customers can handle the whole data side in one place.
MountAIn's customers upload images from their own cameras, and Roboflow's auto-labeling drafts the annotations so people only review and correct them. Each dataset version records its preprocessing, so a team can trace which data trained which model. Hosted training removes the need to set up GPUs. Roboflow's Python package then hands the trained weights to MountAIn Studio, which compiles them for the camera's neural processor and deploys them over USB from the same script.
"Most of our customers don't have a machine learning team," Huard said. "Roboflow lets an embedded engineer label data, train a model and hand it to our compiler in an afternoon, without building a training pipeline of their own."
The Result: Custom Models on a Solar-Powered Camera
In MountAIn's early tests, a YOLO nano model trained in Roboflow runs at up to 22 frames per second on IBEX. MountAIn Studio reports accuracy after each compile stage, so teams see what quantization costs before a model ships. When field data shows a miss, the team adds those frames in Roboflow, retrains, and pushes a new model to the camera without reflashing its firmware.
MountAIn is preparing a fire-detection deployment in Canada and plans to put 1,000 IBEX units in the field by the end of 2027. MountAIn has published a step-by-step guide to deploying Roboflow models on IBEX, and developers can request access to MountAIn Studio through its Early Adopter Program.
MountAIn builds AI deployment infrastructure for constrained hardware. Its MountAIn Core compiler and middleware take trained models onto low-power chips, through MountAIn Ready for evaluating hardware and MountAIn Onboard for adding AI to existing products. Its IBEX camera brings event-level vision to remote, off-grid sites. MountAIn is based in Grenoble, France. Learn more at mountainedge.ai (opens in a new tab).
About Roboflow
Roboflow gives organizations the platform to build visual intelligence that understands and acts in the physical world. The Roboflow platform spans labeling, training, deployment, and applications, and is used by more than one million engineers and over half of the Fortune 100 across manufacturing, logistics, rail, defense, healthcare, retail, and sport. Learn more at roboflow.com.