RegionKit RegionKit Open Editor →
Free · Browser-based · No install

YOLO Annotation Tool Online — Free, No Install

Annotate images for YOLO object detection training directly in your browser. Draw bounding boxes and polygons, assign class labels, and export YOLO TXT format — free, no install, no account, no image upload.

Open RegionKit free

How to get started

  1. 1
    Open RegionKit
    Go to editor.regionkit.app in your browser. No install, no account. Works in Chrome, Firefox, Edge, and Safari.
  2. 2
    Load your image
    Drag an image onto the canvas, use the file picker, or paste from clipboard. For YOLO training data, use the actual training images you want to annotate.
  3. 3
    Draw bounding boxes
    Press R to activate the Rectangle tool. Click to set the first corner, then click again to set the opposite corner. The box is committed immediately.
  4. 4
    Label each annotation
    Select a box and enter its class label in the Properties panel. Use tags to add YOLO class IDs (e.g. tag: "0" for class 0) for clean export mapping.
  5. 5
    Export as YOLO TXT
    Click Export → YOLO TXT. RegionKit outputs one .txt file per annotation in normalised format: class_id x_center y_center width height.
  6. 6
    Use in your training pipeline
    Place the .txt files alongside your images in the standard YOLO directory structure. Point your training config (data.yaml) to the parent directory.

YOLO annotation format

YOLO models use a specific annotation format: one .txt file per image, with one line per annotated object: class_id x_center y_center width height. All five values are space-separated. The class ID is an integer. The four geometry values are floating-point numbers normalised to the image dimensions — values between 0 and 1.

RegionKit exports this format directly. Each Rectangle annotation in your scene becomes one line in the YOLO .txt file, with the normalisation applied automatically. Label-to-class-ID mapping is handled in your data.yaml file.

Bounding boxes and polygons

Bounding boxes (YOLO detection) — Press R for the Rectangle tool. Click to anchor the first corner, click again to commit. Boxes are the standard format for YOLO object detection tasks.

Polygons (YOLO segmentation) — Press P for the Polygon tool. Click each vertex, double-click or press Enter to close. YOLOv8 and later support polygon segmentation masks. RegionKit exports polygons in normalised YOLO segmentation format as well.

Organising annotations by class

RegionKit's layer system maps well to YOLO class organisation. Create one layer per class (e.g. person, car, bicycle), assign each annotation to the appropriate layer, and set the label to match the class name.

In your data.yaml, list the class names in the same order as your layer/label assignments. The class index (0, 1, 2…) is determined by position in the names list.

YOLO directory structure

YOLO training expects images and labels in mirrored directories:

The filename of each .txt must match the corresponding image filename (e.g. img001.txt for img001.jpg). Images with no annotations should have an empty .txt file.

When RegionKit is the right YOLO annotation tool

RegionKit works best when you need to annotate a small batch of images quickly without setting up a server, when Python or Docker isn't available on your machine, when you need to work offline, or when you want zero image upload. For annotating thousands of images with a team, platforms like CVAT or Roboflow are better suited.

Read the full walkthrough in the YOLO annotation tutorial on the RegionKit blog.

Open the YOLO annotation tool →

Try RegionKit — it's free

Browser-based ROI editor. No install, no account, no data upload.

Open the Editor