How do I build a pass/fail check that compares an expected count?

Key facts

What you need

Images of each item and a detection model trained on its parts

What you get

A PASS or FAIL per image, with the count found for each part

Image sources

Line cameras, IP and RTSP cameras, phone photos and saved images

Runs on

Cloud API, or an edge device or server running Docker

How to build a pass/fail count check, step by step

Six steps from images of your product to a PASS or FAIL for every item.

1. 01

   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 images, start training, build the Workflow and run the model from a prompt, at no extra cost beyond your Roboflow plan.

2. 02

   Photograph good and faulty items the way the line will see them

   Capture images from the camera position, distance and lighting you will use in production. Include items with a part missing and items with one part too many, so the model learns from every case the check has to judge.

3. 03

   Label every part, with one class per part type

   Draw one tight box around every part and give each part type its own class, such as screw, clip or bracket. Roboflow Annotate can pre-label images with SAM 3, so you correct boxes instead of drawing each one.

4. 04

   Train a detection model

   Train an object detection model such as RF-DETR on your labeled images with Roboflow Train. A count check is only as good as the model, so test it on images it has not seen before you trust its verdicts.

5. 05

   Compare each part's count with the count you expect

   Run the model, count the detections for each class and compare every count with the number the item should have. The item passes only when every part type matches its expected count.

   PythonCopy

   import supervision as sv
   from collections import Counter
   from inference import get_model
   
   model = get_model("your-project/1")
   expected = {"screw": 4, "clip": 2}
   
   result = model.infer("part_001.jpg", confidence=0.5)[0]
   detections = sv.Detections.from_inference(result)
   counts = Counter(detections.data["class_name"])
   
   passed = all(counts[name] == n for name, n in expected.items())
   print(dict(counts), "PASS" if passed else "FAIL")
   # {'screw': 3, 'clip': 2} FAIL

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

6. 06

   Build the check into a Workflow and deploy it

   Build a Workflow with your model, a Property Definition block that counts the detections and an Expression block that compares the count with an expected-count input and returns PASS or FAIL. Passing the expected count as an input lets one Workflow check several products. Run it on images or a camera stream with Roboflow Deploy in the cloud or on an edge device.

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.