Build with RF-DETR
Build a Part Counting App Free with RF-DETR
Count every bolt, nut, and fastener in a tray, bin, or kit from one image, sorted by type. Label images, train RF-DETR free, and deploy it as a workflow.
- No credit card
- No ML team required
- Edge, on-prem or API
Build a Part Counting App with Roboflow’s Free RF-DETR
In short
A part counting app uses computer vision to find and count small parts like bolts and nuts in an image of a tray, bin, or kit, and returns a separate count for each part type. The model to train for it is RF-DETR, Roboflow's open-source real-time detection model. With Roboflow you label your parts, train RF-DETR for free, and deploy it as a workflow that returns an annotated image and a count for every part type.
How it works
The workflow behind the part counting app
Use this Part Counting with RF-DETR workflow: each image goes to an RF-DETR model that finds every bolt and nut, then through a per-class confidence filter. Two counters tally the bolts and nuts on one track while the other draws boxes and labels, and both meet where the counts are written onto the image. Every block is editable in the low-code builder, and the Roboflow Agent can build it from a description.
- output_image
- bolt_count
- nut_count
- predictions
How to build a part counting app with RF-DETR
Open the workflow and add your images
Click "Use this free template" to copy the workflow into your Roboflow workspace. Upload images of your trays, bins, or kits taken with the camera and lighting at your station, with parts both spread out and piled up.
Label parts and train RF-DETR
Use Auto Label to box each part by type, such as bolt, nut, or washer, then review and correct the results. Train an RF-DETR model and select it in the detect_parts block. RF-DETR fine-tunes well on small custom datasets, so you can start with the images you have.
Tune the thresholds and add part types
Set the per-part thresholds in filter_by_confidence. For each new part type you label, add a counter like count_bolts so the workflow returns a count for it.
Deploy and connect it to your line
Deploy the workflow as a Serverless Cloud API, or on your own hardware through Inference. Each image returns output_image, bolt_count, nut_count, and predictions for your production system, inventory system, or kit check to use.
Roboflow MCP
Build it from your AI assistant with the Roboflow MCP
Connect the Roboflow MCP Server to Claude, Cursor, or any MCP-compatible assistant and build the whole part counting app in natural language, without leaving your editor.
Ask an assistant like Codex or ChatGPT to search Roboflow Universe for fastener and part datasets, upload and auto-label your tray images, and train an RF-DETR model on bolt and nut classes. It can then create and run the part counting workflow on test images and deploy it to your devices.
Paste it into your coding agent and it walks you through the rest. Setting up by hand? The server is https://mcp.roboflow.com/mcp.
Open the Roboflow listing, select Connect, then sign in to Roboflow. Available on desktop, web and mobile.
Run these commands in your terminal, then sign in to Roboflow. Claude Code CLI has no documented browser install link.
Opens the Roboflow plugin listing. Sign in if prompted and confirm the install on Cursor to add MCP and skills.
Opens the Roboflow plugin listing. Install the plugin and sign in to Roboflow when prompted to connect your account.
Opens the Roboflow plugin listing shared by ChatGPT and Codex. Install the plugin and sign in to Roboflow when prompted to connect your account.
Then run /mcp auth roboflow in Gemini CLI to sign in. Gemini CLI has no documented browser install link.
What your part counting app can do
Trained on your images, an RF-DETR model handles the specific, hard-to-catch cases that matter on your line:
- Count small parts that touch or overlap in a tray or bin
- Tell similar parts apart, like bolts, nuts, and washers
- Keep counting through glare on shiny metal parts
- Count parts lying at any angle or orientation
- Count a full kit from one overhead image
- Add new part types as your bill of materials changes
Want to customize your part counting app?
Describe the part counting app you want in plain English: the parts, where they sit, and what count you need. The Roboflow Agent finds the data, trains an RF-DETR model, and assembles the workflow for you.
Open source
Roboflow’s RF-DETR is open source, Apache-2.0, and free to use
RF-DETR is an open-source, Apache-2.0 licensed real-time object detection model developed by Roboflow. The rfdetr Python package and most RF-DETR models are free to use, including in commercial products, and you can train and run everything on your own hardware. No account or lock-in required.
Apache-2.0 and free
The rfdetr package, Nano to Large detection, every segmentation size and Keypoint are Apache-2.0 licensed and free to use, including commercially.
Install and train locally
pip install rfdetr on Python 3.10+. Fine-tune on your own data in a few lines and run it entirely on your machine.
State of the art on COCO
RF-DETR is a real-time detector that is SOTA on COCO. RF-DETR-Large reaches 56.5 AP at 6.8 ms on an NVIDIA T4, and uses a DINOv2 backbone.
Deploy anywhere
Export to ONNX or TensorRT and run on GPU, CPU, or NVIDIA Jetson at the edge, or serve it through the Roboflow hosted API.
| Models | Licence |
|---|---|
| Detection: Nano, Small, Medium, Large | Apache-2.0 |
| Segmentation: every size, Nano to 2XLarge | Apache-2.0 |
| Keypoint (preview) | Apache-2.0 |
| Detection: Atto, Femto, Pico, XL, 2XL (rfdetr[plus]) | PML 1.0: commercial use included on Roboflow; commercial self-hosting needs the Enterprise add-on |
RF-DETR vs. other free detectors
A part counting app has to find many small, similar parts in one image, often touching or piled together, so the model needs to be both accurate and fast. Here is how RF-DETR, fine-tuned on your own part images, compares for real-time counting.
| Model | Tasks | Best for | License |
|---|---|---|---|
| RF-DETR | Detection, segmentation (keypoints in preview) | Real-time detection you fine-tune on your own data, one model across tasks | Apache-2.0 (Nano to Large, all segmentation); PML 1.0 (Atto, Femto, Pico, XL, 2XL) |
| D-FINE | Detection | Precision localization when you do not need to fine-tune | Apache-2.0 |
| YOLO | Detection, segmentation, pose | Fast CNN-based detection with a large ecosystem | AGPL-3.0 (commercial license required for closed-source use) |
| GroundingDINO | Open-vocabulary detection | Finding objects by text prompt with no training, not real-time | Apache-2.0 |
| SAM 2 / SAM 3 | Segmentation | Pixel-precise masks; SAM 3 adds text-prompted segmentation | Apache-2.0 (SAM 2), SAM License (SAM 3) |
| Florence-2 | Vision-language | Reading, captioning, and OCR on images | MIT |
The takeaway for this app: RF-DETR runs in real time, so it can count parts as fast as trays and kits move through your station. It fine-tunes well on small custom datasets and is open source, so you can train on your own parts and run the model wherever you need it.
Build your part counting app free today
Label, train, and deploy your part counting app in one place.