⭐ Featured
JSON Formatter
Format JSON instantly with indentation controls, validation feedback, and clean readable output.
●Developer Tools
🔒 Local⚡ Free
Draw, edit, and inspect image annotations with interactive canvas and COCO, LabelMe, and YOLO import/export support.
Draw annotations (bounding boxes, polygons, circles, lines) on images for machine learning datasets (like window, door, facade, or object detection training datasets). Manage classes, and import/export datasets in COCO, LabelMe, and YOLO formats with a real-time synchronized JSON editor.
Designed for direct use without extra steps or inflated trust claims.
Upload or drag-and-drop an image or PDF to begin
Universal Annotation Tool & Dataset Builder is an offline-first developer utility for code inspection, encoding, and formatting. Draw, edit, and inspect image annotations with interactive canvas and COCO, LabelMe, and YOLO import/export support. Designed for engineers and system administrators, all syntax operations, parsing routines, and transformations run inside the local JavaScript runtime, keeping production payloads, API keys, and database dumps completely private. Built for zero-latency execution, Universal Annotation Tool & Dataset Builder eliminates network round-trips by computing results directly within your device session. Whether you are running routine image annotation tool workflows or fine-tuning output parameters for coco dataset builder, all data remains isolated in volatile memory with instant feedback.
Step 1: Upload Image or PDF
Select and load your target image file or multi-page PDF document. Pages will load into the preview slider below the canvas.
Step 2: Select Class Label and Draw Shapes
Create classification classes and use the square, circle, triangle, line, polygon, or point marker drawing tools to outline objects directly on the canvas.
Step 3: Refactor and Apply Metadata
Add custom field schemas (like difficulty, occlusions, tags) to shapes. Use the JSON editor drawer to refactor class names, offsets, or coordinate scale factor.
Step 4: Export Dataset Files
Select your target computer vision formatting (COCO JSON, LabelMe JSON, or YOLO text) and click to trigger local download of your dataset files.
No. The entire process (file rendering, canvas calculations, and dataset formatting exports) is executed 100% locally in your browser. None of your data ever leaves your computer.
Yes. You can import existing annotations in COCO JSON, LabelMe JSON, or YOLO text formats by drag-dropping the JSON/text files to edit them.
When a multi-page PDF is loaded, it renders each page as a canvas background. You can select pages using the preview thumbnails below the canvas to manage drawings for each page independently.
Yes. All active shapes feature standard resize and rotate anchors on the canvas. You can toggle layer visibility and lock attributes to prevent accidental modifications.
You can use rectangles, circles, triangles, lines, points, custom freehand loops, and arrow shapes to annotate your images or documents.
Yes. Holding down the Shift key while dragging any of the transformer anchors locks the aspect ratio of the bounding boxes.
You can select any annotation and use the Arrow keys on your keyboard to nudge it by 1 pixel, or hold Shift + Arrow keys to nudge it by 10 pixels for precise placement.
Yes. Under the Props tab in the right-hand panel, you can add custom key-value metadata fields (such as occlusions, tags, or notes) that will be bundled into exported JSON formats.
Focus Mode expands the workspace into a clean, fullscreen interface to maximize screen real estate and remove distraction during large annotation tasks.
Yes. The workspace automatically saves your shapes, categories, and settings locally every 800 milliseconds, allowing you to resume your session if the browser refreshes.
Engineered for high-security, client-side browser execution with zero network transmission.
Client V8 & WASM
Computed in local browser memory with zero upload lag.
0 Bytes Transmitted
No payload uploads, database logs, or external API relays.
Volatile RAM Only
Inputs automatically purged when the page or tab is closed.
Clean & Unwatermarked
Native format downloads and clipboard synchronization.
Need alternatives or related utilities? Explore more in Annotation Tools or view the complete tools directory.
This tool processes all inputs locally in your browser. Your files and data are never uploaded to our servers, keeping them 100% private.
Describing the keyboard-driven autocomplete and scored ranking algorithm we deployed. Details of scoring prioritizations, highlight text, and Arrow key listeners.
How local ML dataset tools like YOLO bounding box creators and COCO JSON viewers run directly inside browser sandboxes using client-side vector canvases.