YOLO-Label
Qt desktop tool for drawing YOLO bounding boxes, with optional ONNX model auto-labeling.
About
Written in C++ with Qt, this desktop annotator exists because its author found older box-drawing tools tedious while preparing a YOLOv3 training set. Instead of dragging, you click once at one corner and again at the opposite one, which is gentler on the wrist, then move through images with A and D or the mouse wheel, saving as you go. Each image gets a plain text file in YOLO format with a class index and normalized box coordinates, and class names come from a .names or .txt list. Version 3 added auto-labeling: load an Ultralytics detection model exported to ONNX (YOLOv5, YOLOv8, YOLO11, YOLO12 and later, including NMS-free exports) and it proposes boxes for one picture or the whole folder, reading class names from the model metadata. An optional link to a hosted open-vocabulary detection service needs an account and API key. Extras include undo and redo, copying boxes between images, keyboard nudging and resizing, zoom up to 10x and a contrast slider. Prebuilt binaries cover Windows x64, a Linux AppImage and Apple Silicon macOS; building from source needs Qt 6. MIT licensed, so it suits commercial dataset work too.
Should you use YOLO-Label?
Pick it when
Pick this when you only need bounding boxes for a YOLO object detector and want a lightweight desktop tool with drag-to-draw boxes, keyboard class switching, and YOLO text files written directly with no conversion step.
Look elsewhere when
Skip it for polygons, masks, keypoints, or any segmentation task, or when you want model pre-labels; X-AnyLabeling covers both and still exports YOLO format, while Labelme handles polygons under MIT.
Alternatives to YOLO-Label
- X-AnyLabeling
Also exports YOLO format and adds YOLO and SAM auto-labeling plus polygons and keypoints, at the cost of a heavier GPL-3.0 app.
- Labelme
Supports polygons, circles, and lines for segmentation, but writes its own JSON that must be converted to YOLO format.
- VGG Image Annotator
Runs in a browser with no install and supports more shapes and attributes, but its exports need converting before YOLO training.
- Autodistill
Skips manual boxes by auto-labeling with foundation models and training YOLOv8, when your classes are common enough to describe in a prompt.
Reviews (0)
Leave a Review
No reviews yet. Be the first to review!
Details
- Category
- Data Labeling & Annotation
- Price
- Free
- Platform
- Local/Desktop
- Difficulty
- Beginner (1/5)
- License
- MIT
- Added
- Apr 3, 2026