One project, from the first image to a running system.
Detection software, camera integration and managed services are not separate products. We deliver them together, as one project built around your process.
Models trained on your process.
CAM-AI is trainable detection software. We train the models on images from your own cameras, for the categories that matter in your process: a foreign object, a tilted part, an empty fixture. Machine-specific plugins connect the results to the systems you already run.
Every result is a confidence score between 0 and 1 for each trained category. What happens with it is decided in your systems.
How the technology works- Models for your categoriesTrained on images from your process, for exactly the errors that matter to you.
- Machine-specific pluginsDetection logic built for the tool or line it runs on.
- A score for every frameA confidence value per trained category, in real time.
- Output via APIJSON, SQL, HTTP, e-mail and others via plugin, into your control, MES or alarm systems.
- Alarm rules tailored to your processRules that combine categories and cameras, defined together with you. What follows is your decision.
Cameras where the error first becomes visible.
Detection is only as good as the image. We plan the camera positions, align them precisely, also in tight machine spaces, and support the installation on your network and your systems. Integration is part of every CAM-AI project, not a separate installation service.
- Camera planning
- Where the error first becomes visible, and from which angle.
- Precise alignment
- Also inside tight machine spaces and process tools.
- Installation support
- We plan the installation and guide it on site, on your network and your hardware.
- System connection
- Results to your MES, control or alarm systems via plugin or API.
- Camera advice
- Existing IP cameras where they are good enough, high-end industrial cameras and optics where your process needs more.
Models get better with operation. We keep them that way.
Processes, parts and light change over time. Over a service period, we keep training the models on data from live operation, review how they perform and keep the software current. You talk to the people who build the system.
- Continued training and refinement of the models on data from live operation
- Regular performance reviews, with a recommendation when a threshold should change
- Software updates for the CAM-AI core and the plugins in use
- Direct support from the people who build the system
Start small. Decide at every step.
Every project runs in three phases. After each one you see results and decide whether to go on.
- 01Free
Proof of concept
You can detect what you can see, and the proof of concept shows it on images or videos from your own process. CAM-AI learns fast: in a first proof of concept, the models reached 96 % average certainty after a single training session.
- You bring
- A clear visual question and, after a short call, images or videos from the point you want to watch.
- You get
- A first trained model and an honest view of what a camera can see there.
- Then
- You decide whether to go on. No obligation.
- 02Priced once
Set-up & engineering
We plan the cameras and support their installation, train the models on live images and connect the results to your systems.
- You bring
- Access to the line or tool, and a contact in automation or IT.
- You get
- A running detection system, integrated into your process.
- Then
- Before go-live, measured performance and a recommended threshold for you to confirm.
- 03Service period
Managed services
We keep the models trained, the software current and stay your direct contact.
- You bring
- Feedback from operation: what was right, what was missed.
- You get
- Models that improve with operation, threshold reviews, updates and support.
- Then
- Runs over a service period that scales with the number of cameras and trained models.
Pricing tailored to your project. Set-up is priced once for your process. Managed services scale with the number of cameras and trained models.
Request a demoWe measure. You decide.
CAM-AI tells you how likely it is that something is there. Your process knows what that should mean. That is why the decision stays with you.
- CameraImages from your processEvery frame, at the point where the error becomes visible.
- CAM-AIA confidence score per categoryBetween 0 and 1, in real time, handed to your systems via plugin or API.
- Your systemsYour threshold, your actionAn alert, a notification or a step in your process. You set it, in the systems you already run.
Measured model performance on your process: how many false alarms and how many missed cases to expect at each threshold, and a recommended value with the reasoning behind it.
The threshold that goes into your system and the action it triggers. A false alarm and a missed object cost different things in every process, and you know which weighs more in yours.
Before the first call.
- Do we need new cameras?
- Not necessarily. Standard IP cameras via RTSP, RTMP, MJPEG and other standard streams work. Where your process needs more, we advise on industrial cameras and optics.
- Where are our images processed?
- Training and detection run on servers inside your facility. Production images can stay where they were taken.
- What does the proof of concept cost?
- Nothing. After a short call, you send us images or videos from your process, we train a first model and show you the results. If they convince you, we propose a set-up for your process. If not, it ends there.
- Is CAM-AI open source?
- The core is, on GitHub: auditable and without vendor lock-in. Machine-specific plugins, integration and managed services are what we deliver in a project.
You can detect what you can see.
Every project starts with one question: what do you want to detect? Tell us, and we look at your process together.
Request a demo