Put Roboflow to work
on your existing cameras.

Connect your camera to Angelcam

Add your camera in MyAngelcam and check that its live view works. For a prototype using your own cameras, create a Personal Access Token. Applications accessing customers’ cameras should use OAuth2.

Keep credentials in your service’s environment or secret store.

Get a fresh camera image

Use Angelcam’s documented camera endpoints to retrieve an image. Check the snapshot’s created_at value: a regular preview can be up to 24 hours old. For a current check, use a live snapshot and record its capture time.

For your first test, save the JPEG as snapshot.jpg. Open it and confirm the objects you want to detect are visible.

Build your Roboflow Workflow

In your workspace, open Workflows → Create Workflow. Add an image input named image and an Object Detection Model block. Select a pretrained model for supported classes, or use your own trained model.

Add a count or visualization if useful, then expose the predictions you need as outputs. Save and test with your camera image. Publish the Workflow version you intend to call.

Send the image to Roboflow

Open Deploy in the Workflow editor and use its Python example. Copy the workspace and Workflow identifiers. The example below sends a local image to Roboflow’s hosted service.

Set ROBOFLOW_API_KEY, ROBOFLOW_WORKSPACE, and ROBOFLOW_WORKFLOW_ID in your service’s environment. This example expects an input named image; match your own Workflow’s input name.

import json
import os
from pathlib import Path
from inference_sdk import InferenceHTTPClient, InferenceConfiguration

# Supply a fresh Angelcam image saved by your camera-fetching service.
image = Path(os.getenv("SNAPSHOT_PATH", "snapshot.jpg"))
if not image.is_file():
    raise SystemExit(f"Snapshot file not found: {image}")

client = InferenceHTTPClient(
    api_url="https://serverless.roboflow.com",
    api_key=os.environ["ROBOFLOW_API_KEY"],
).configure(InferenceConfiguration(api_key_transport="header"))

result = client.run_workflow(
    workspace_name=os.environ["ROBOFLOW_WORKSPACE"],
    workflow_id=os.environ["ROBOFLOW_WORKFLOW_ID"],
    images={"image": str(image)},
)
print(json.dumps(result, indent=2))

Download the example · Workflow deployment documentation

The example analyzes an image you supply. Your camera-fetching service completes the connection to Angelcam.

Check the results on your own scene

Compare predictions with the image. Test different lighting, empty scenes, overlapping objects, and the smallest objects you need to detect. Tune the confidence threshold against missed detections and false positives.

Define the output your application needs: a per-image count, objects inside a region, or a flagged image for review. A confidence score is a model score, not a guarantee of accuracy.

Add a schedule and a useful action

Schedule your service to fetch a fresh image and call the Workflow. Store the camera ID, capture time, analysis time, Workflow version, and predictions. Then add a dashboard update or notification rule.

Handle offline cameras, stale images, failed requests, and repeated detections. Use a notification cooldown so the same object does not generate an alert on every run.

Illustrative workload for continuous 24-hour scheduling. Configure the real schedule in your service. This page does not activate camera access or run inference.