Sponsors

• AIM(https://www.aimdefence.com/)

WHAT IS THIS?!

Reinventing The Wheel?!!!!

In the world, there are many good image-labeling tools for object detection. -e.g. , (Yolo_mark, BBox-Label-Tool, labelImg).

But... I've reinvented one...

WHY DID YOU REINVENT THE WHEEL? ARE YOU STUPID?

When I used the pre-existing programs to annotate a training set for YOLO V3, I was sooooooooooo bored...

So I thought why it is so boring??

And I found an answer.

The answer is that pre-existing programs are not sensitive.

So I decided to make a sensitive image-labeling tool for object detection.

SHOW ME YOUR SENSITIVE IMAGE-LABELING TOOL!!

It's the SENSITIVE image-labeling tool for object detection!

https://user-images.githubusercontent.com/35001605/211560039-367f27d7-63ab-4342-824e-9f47f2afbc35.mp4

HMM... I SAW THIS DESIGN SOMEWHERE

I refer to the website of Joseph Redmon who invented the YOLO.

TUTORIAL / USAGE

Download

Pre-built binaries are available on the Releases page.

OS	Download	Note
Windows (x64)	YoloLabel-Windows-x64.zip	Unzip and run YoloLabel.exe
Linux (x64)	YoloLabel-Linux-x64.AppImage	chmod +x and run
macOS (Apple Silicon)	YoloLabel-macOS.dmg	Open DMG and drag to Applications

Install and Run

For Windows

1. Download YoloLabel-Windows-x64.zip

2. Unzip

3. Run YoloLabel.exe

For Linux

1. Download YoloLabel-Linux-x64.AppImage

2. Make executable and run

chmod +x YoloLabel-Linux-x64.AppImage
./YoloLabel-Linux-x64.AppImage

For macOS

1. Download YoloLabel-macOS.dmg

2. Open the DMG and drag YoloLabel.app to Applications

3. Launch YoloLabel from Applications

Build from source: If you prefer, install Qt 6 (brew install qt@6), clone this repo, then run qmake && make. See Build with ONNX Runtime for auto-label support.

Prepare Custom Dataset and Load

1. Put your .jpg, .png -images into a directory (In this tutorial I will use the Kangarooo and the Raccoon Images. These images are in the 'Samples' folder.)

2. Put the names of the objects, each name on a separate line and save the file( .txt, .names).

3. Run Yolo Label!

4. Click the button 'Open Files' and open the folder with the images and the file(''.names or ''.txt) with the names of the objects.

5. And... Label!... Welcome to Hell... I really hate this work in the world.

This program has adopted a different labeling method from other programs that adopt "drag and drop" method.

To minimize wrist strain when labeling, I adopted the method "twice left button click" method more convenient than

"drag and drop" method.

drag and drop

twice left button click

6. End

USAGE AND OPTIONS

./YoloLabel [dataset dir] [class file] [model.onnx]
# Examples
./YoloLabel ../project/dataset/objects/frames ../project/dataset/objects/obj_names.txt
./YoloLabel ../project/dataset/objects/frames ../project/dataset/objects/obj_names.txt yolov8n.onnx
./YoloLabel ../project/dataset/objects/frames yolov8n.onnx

Arguments are detected by file extension — .onnx files are loaded as YOLO models, all other files are loaded as class name lists. When a model with embedded class names is loaded without a class file, class names are populated from the model automatically.

SHORTCUTS

Key	Action
A	Save and Prev Image
D, Space	Save and Next Image
S	Next Label
W	Prev Label
O	Open Files
V	Visualize Class Name
Ctrl + S	Save
Ctrl + Delete (Windows/Linux) / Cmd + Delete (macOS)	Delete all existing bounding boxes in the image
Ctrl + D, Delete	Delete current image
` (Backtick)	Select first class (class 0)
0-9	Quick select class by number
Arrow Keys	Nudge the bounding box under the cursor (~1-2px step)
Shift + Arrow Keys	Nudge the bounding box under the cursor (~5px step)
Ctrl + Arrow Keys (Windows/Linux) / Cmd + Arrow Keys (macOS)	Resize the bounding box under the cursor (~1-2px step, top-left corner fixed)
Ctrl + Shift + Arrow Keys (Windows/Linux) / Cmd + Shift + Arrow Keys (macOS)	Resize the bounding box under the cursor (~5px step, top-left corner fixed)
Ctrl + C (Windows/Linux) / Cmd + C (macOS)	Copy bounding boxes to clipboard
Ctrl + V (Windows/Linux) / Cmd + V (macOS)	Paste bounding boxes from clipboard
Ctrl + Z (Windows/Linux) / Cmd + Z (macOS)	Undo last action (add, remove, or clear all)
Ctrl + Y (Windows) / Ctrl + Shift + Z (Linux) / Cmd + Shift + Z (macOS)	Redo last undone action
R	Auto Label current image (requires loaded ONNX model)
Ctrl + 0 (Windows/Linux) / Cmd + 0 (macOS)	Reset zoom to 100%

Mouse	Action
Right Click	Delete Focused Bounding Box in the image
Left Click + Drag on existing box	Move/reposition the bounding box
Option + Left Click (macOS) / Alt + Left Click (Win/Linux)	Change class of focused bounding box to the currently selected label (no need to remove and redraw)
Double Click on color column in label table	Change label color
Ctrl + Scroll (Windows/Linux) / Cmd + Scroll (macOS)	Zoom in/out (centered on cursor, up to 10x)
Ctrl + Left Drag (Windows/Linux) / Cmd + Left Drag (macOS) or Middle Mouse Drag	Pan while zoomed in
Wheel Down (when cursor is over image)	Save and Next Image
Wheel Up (when cursor is over image)	Save and Prev Image

Auto-Label (Pseudo Labeling)

Tired of drawing every single bounding box by hand? Load a pre-trained YOLO model and let it do the boring work for you.

YOLO-Label supports local ONNX inference — just export any Ultralytics detection model to .onnx and load it. Class names, input size, and model configuration are all read from the ONNX metadata automatically. No separate config files needed.

Supported Models

Any Ultralytics detection model exported with model.export(format="onnx"):

Model	Note
YOLOv5	Anchor-based, objectness score
YOLOv8, YOLO11, YOLO12, YOLOv26	Anchor-free
End-to-end models	NMS baked in

How to Use

1. Open your dataset as usual (images + class file)
2. Click Load Model and select a .onnx file
3. Adjust the confidence threshold with the slider (default 25%)
4. Click Auto Label (or press R) to detect objects on the current image
5. Click Auto Label All to batch-process all images
6. Review and correct results using the existing manual annotation tools — auto-labeled boxes can be moved, deleted, or undone (Ctrl+Z)

Tip: You can skip the class file entirely. If the model has embedded class names (all Ultralytics exports do), they will be loaded automatically.

Build with ONNX Runtime

Pre-built releases include ONNX Runtime. To build from source with auto-label support:

# Download ONNX Runtime (one-time setup)
./scripts/download_onnxruntime.sh

# Build
qmake YoloLabel.pro "ONNXRUNTIME_DIR=$PWD/onnxruntime"
make -j$(nproc)

Without ONNX Runtime, the app builds and works normally — just without the auto-label feature.

Cloud Auto-Label (yololabel.com)

No local GPU? No problem. YOLO-Label integrates with yololabel.com — a cloud inference service that runs open-vocabulary object detection on your images without requiring any local model or GPU.

Create an Account and Get an API Key

1. Go to yololabel.com and sign up with email + password or Sign in with Google
2. If you registered with email, check your inbox and click the verification link
3. Log in and navigate to API Keys, then click Create Key
4. Give the key a name and copy it immediately — it is shown only once

How to Use

1. Open your dataset as usual
2. In the ⚙ AI Settings tab on the right panel, paste your API key and (optionally) a custom detection prompt
   • The prompt is a semicolon-separated list of labels, e.g. car; person; bicycle
   • Leave blank to use the loaded class names automatically
3. Click ☁ Auto Label AI to detect objects in the current image
4. Click ☁ Auto Label All AI to process the entire dataset

Batch Processing

When ☁ Auto Label All AI is clicked with multiple images, images are submitted in batches of up to 20 per request using the /v1/jobs/batch endpoint. Progress is shown live in the button label.

Contrast Adjustment

Use the Contrast slider at the top of the window to adjust image brightness/contrast in real-time. This is useful when labeling dark or overexposed images. The slider ranges from 0% to 100% (default 50%).

Usage Timer

A timer in the status bar counts how many hours (and minutes/seconds) you have been using the program. It runs only while the window is focused (switches to another app to pause). Use the Reset button in the status bar to zero the timer at any time.

Button Events

Remove

It was replaced by the shortcut Ctrl + D.

ETC

You can access all image by moving horizontal slider bar. But when you control horizontal slider bar, the last processed image will not be saved automatically. So if you want not to lose your work, you should save before moving the horizontal slider bar.

CONCLUSIONS

I've reinvented the wheel.

TO DO LISTS

• Upload binary file for easy usage for windows and ubuntu
• deployment for ubuntu
• macOS Developer signing for Gatekeeper