PixelTagger is the desktop-native image annotation tool built for machine learning engineers who need to build accurate computer vision datasets — without cloud uploads, subscriptions, or slow web interfaces.
Focused, keyboard-first, and offline. PixelTagger removes every obstacle between you and a fully labeled dataset.
Navigate, annotate, and switch images without touching the mouse. Annotation rhythm that doesn't break.
No uploads. No accounts. No privacy risk. Your images and labels never leave your machine.
YOLO, COCO JSON, DOTA, CSV, VOC XML. Switch output formats anytime — annotations are never locked in.
Pull individual frames from any video file directly into your dataset with the built-in frame picker.
Run any YOLO model (v10/v11/v12) on a video and automatically extract only the frames where your labels appear — or don't appear (inverse mode).
One-click randomized split of images and annotations into Train and Test subsets at any ratio.
Run YOLO inference on every frame of a video and save a new annotated video with bounding boxes, keypoints, and labels rendered on each frame.
Lock zoom to annotated regions. Automatically zooms in when switching images — ideal for small objects.
Generate a visual grid of annotated crops for fast QA review of your labeled dataset.
From simple object detection to 3D cuboid annotation — one tool handles every computer vision labeling task.
Bounding Box
Polygon
Keypoint
Line
Rotated Box
Cuboid
Ellipse
Run the application installer. One click, no configuration needed.
Select your images folder and an annotations folder — it can be empty.
Draw shapes, assign labels. Annotations save automatically as you go.
Export to YOLO, COCO, DOTA or CSV — plug directly into your pipeline.
PixelTagger reads and writes the formats your ML frameworks already expect. No conversion scripts needed.
All annotation shapes, all export formats, AI-powered video tools — in one package.
Native Windows installer. Run the .exe and launch from Start Menu.
DOWNLOAD for Windows (.exe)Requires Windows 10 or 11 (64-bit)
DMG disk image. Drag PixelTagger to Applications, then open from there.
DOWNLOAD for macOS (.dmg)Requires macOS 11 Big Sur or later
Self-contained AppImage. No installation needed — just chmod +x and run.
DOWNLOAD for Linux (.AppImage)Requires Ubuntu 20.04 or later
⚠️ Beta software — some features may be incomplete. Report a bug →