Computer vision answers · Energy

How do I detect utility poles and their components with computer vision?

Without a crew climbing, or driving past, every pole to look.

Short answer

Train an object detection model to find each utility pole and its components, such as crossarms, insulators and transformers, with Roboflow for free. Then assign each component to the pole it sits on, so every pole in your footage comes with its own list of parts. Label the whole pole as its own class as well as each component, so components from a pole in the background never get counted on the pole in front.

Start free
Roboflow agent · free to start

Skip the setup. Describe your images.

Describe your images and what you need to find. The agent sets up labeling, training and deployment with you.

Key facts

What you need
Pole photos or video and a model trained on your components
What you get
Every pole found, with a list of the components on it
Image sources
Drone, vehicle-mounted and handheld cameras
Runs on
Cloud API, or an edge device or server running Docker
Utility pole image with object detection labels for the pole, crossarm, 3 insulators, and a transformer. A legend lists these components with their counts.

How to build utility pole detection, step by step

Six steps from drone or vehicle footage to every pole with its components listed.

  1. Connect Roboflow to your coding agent, ChatGPT or Codex

    Add the Roboflow MCP server to Claude, Cursor or Gemini CLI, or install the Roboflow plugin in ChatGPT or Codex. Your assistant can then upload and auto-label inspection images, start training, build the Workflow and run the model from a prompt, at no extra cost beyond your Roboflow plan.

  2. Collect footage the way you inspect

    Gather images or video from the cameras you will use in the field, whether drone, vehicle-mounted or handheld. Include poles close up and in the distance, in the light and weather your crews work in.

  3. Label each pole and each component as its own class

    Draw a box around each whole pole and around every crossarm, insulator and transformer on it. Roboflow Annotate can pre-label images with SAM 3 from plain-language prompts, and public pole datasets on Roboflow Universe can give you a head start.

  4. Train a detection model

    Train an object detection model such as RF-DETR on your labeled images with Roboflow Train, and test it on a route it has not seen.

  5. List the components on each pole

    Run the model, split the detections into poles and components, and give each component to the pole whose box contains the component's center point.

    Python
    import supervision as sv
from inference import get_model

model = get_model("your-project/1")
result = model.infer("route_12_frame.jpg", confidence=0.4)[0]
detections = sv.Detections.from_inference(result)

poles = detections[detections.data["class_name"] == "pole"]
parts = detections[detections.data["class_name"] != "pole"]
centers = parts.get_anchors_coordinates(sv.Position.CENTER)

for i, (x1, y1, x2, y2) in enumerate(poles.xyxy):
    inside = (centers[:, 0] >= x1) & (centers[:, 0] <= x2) & (centers[:, 1] >= y1) & (centers[:, 1] <= y2)
    print(f"pole {i}:", parts.data["class_name"][inside].tolist())
# pole 0: ['crossarm', 'insulator', 'insulator', 'insulator', 'transformer']

    Rather not write code? Ask the Roboflow agent to set this up for you

  6. Run every inspection and send findings to work orders

    Build a Workflow that runs the model on live, recorded or local footage with Roboflow Deploy, and pass each finding on to your crews, for example as a work order in Salesforce.

Why 2 million engineers build on Roboflow

  • Label faster

    Auto-label with SAM 3 or your own model, then fix the misses instead of drawing every box.

  • RF-DETR

    Roboflow’s open-source real-time detection model, built for accuracy (opens in a new tab) on small objects.

  • Logic built in

    Counting, zones, tracking and alerts are ready-made Workflow blocks, so there is less code to write.

  • Your model, your weights

    Use your trained weights on every plan, including free, through the open-source Inference package.

  • Run it anywhere

    Self-host on a workstation, Docker, an on-prem GPU or an edge device, or call the cloud API.