Alert on detections

You want to be notified when your camera detects something specific – a person in a restricted area, a missing hard hat, or an anomaly on a production line. This guide shows you how to connect your vision service to Viam’s trigger system so you receive an email or webhook whenever a detection occurs. No custom code is required.

Concepts

The alert pipeline

The alert system chains three resources together:

1. Filtered camera – a camera module that only passes images to the data management service when specific detections or classifications are present.
2. Data management service – captures images from the filtered camera and syncs them to the Viam cloud.
3. Trigger – fires when new data syncs, sending an email or webhook notification.

Because the filtered camera only passes images that match your criteria, every synced image represents a detection event. The trigger fires on each sync, turning data events into alerts.

Filtered camera behavior

The filtered-camera module wraps an existing camera and applies a vision service as a filter. It functions as a normal camera for live viewing and API calls. The filtering only affects what gets captured by the data management service.

You configure the filter with:

• A label – the detection or classification class name to look for (such as “Person” or “NO-Hardhat”).
• A confidence threshold (0.0-1.0) – the minimum confidence score required. Only images where the model meets this threshold are captured.

Alert frequency

Triggers support configurable notification frequency to prevent alert fatigue. For example, you can limit alerts to a maximum of one per hour even if detections occur continuously.

Steps

1. Configure a filtered camera

Add the filtered camera module to your machine:

1. Navigate to your machine’s CONFIGURE tab.

2. Click + and select Configuration block.

3. Search for filtered-camera and select it.

4. Click Add component, name it objectfilter-cam, and click Add component again to confirm.

5. Add configuration attributes:

   • Template
   • Example: Hard hat detection

   Replace <camera_name> and <vision_service_name> with the names of your camera and vision service. Choose either objects (for bounding-box detections) or classifications (for image-level labels) and remove the other.

   {
     "camera": "<camera_name>",
     "vision_services": [
       {
         "vision": "<vision_service_name>",
         "classifications": { "<label>": 0.5 },
         "objects": { "<label>": 0.5 }
       }
     ]
   }
   

   This example uses a YOLOv8 model named yolo to detect workers without hard hats:

   {
     "camera": "my_webcam",
     "vision_services": [
       {
         "vision": "yolo",
         "objects": {
           "NO-Hardhat": 0.6
         }
       }
     ]
   }
   

6. Click Save.

For more details, see the filtered-camera module README.

2. Configure data capture and sync

Add the data management service to capture and sync filtered images:

1. Click + and select Configuration block.
2. Search for data management and select it.
3. Click Add component, name it data-manager, and click Add component again to confirm.
4. Leave the default attributes and click Save.

Enable data capture on the filtered camera:

1. Locate the objectfilter-cam panel.
2. Click the Data Capture button.
3. Select GetImages as the type.
4. Set the capture frequency to 0.2 images per second (one image every 5 seconds). Adjust as needed for your use case.

3. Configure a trigger

Add a trigger to send alerts when filtered images sync:

1. Click + in the left sidebar and select Trigger.

2. Enter a name and click Create.

3. Set Type to Data has been synced to the cloud.

4. Set Data Types to Binary (image).

   

5. Add notification methods:

   Email specific addresses: Toggle on, add email addresses, and set the alert frequency.

   Email all machine owners: Toggle on and set the alert frequency.

   Webhook: Click Add Webhook, enter the URL of your cloud function, and implement logic to process the webhook payload. Use this to integrate with external services like Twilio, PagerDuty, or Zapier.

6. Set the notification frequency (for example, maximum one alert per hour).

7. Click Save.

Try It

1. Point your camera at an object your model recognizes and wait for the capture interval to pass.
2. Check the TEST panel on your vision service to confirm detections are occurring with sufficient confidence.
3. Navigate to the DATA tab and verify that images are syncing.
4. Check your email or webhook endpoint for the alert.

Troubleshooting

What’s Next

• Act on Detections – build a module that responds to vision results in real time.
• Triggers Reference – full documentation for trigger types, webhook payloads, and configuration options.
• Capture and Sync Data – learn more about configuring data capture and sync intervals.