YOLO BBox Annotation Tool - https://lddl.github.io/yolo-ann

Fork of https://github.com/drainingsun/ybat

Fast and efficient BBox annotation for your images in YOLO, and now, VOC/COCO formats!

USAGE

Online

1. Navigate to https://lddl.github.io/yolo-ann/
2. Load images and classes and start bboxing!

Offline

Click to expand

1. Download repository

   git clone https://github.com/LdDl/yolo-ann.git

2. Navigate to root folder

   cd yolo-ann

3. Open index.html in your browser.

4. Load images and classes and start bboxing!

5. Optional configuration:

   5.1. Open index.js.

   5.2. Edit section named parameters.

UI should look something like this:

Differences from original repository

Main differences are:

• Using FabricJS instead of unnamed one. Tons of changes have been done, if you find a bug, please open an issue or open a pull request.
• Updated version of jszip
• Updated version of FileSaver.js
• Panning is done via wheel button
• Accept *.names, *.csv and *.txt extensions for classnames file

Compatibility

All browsers that support ES6 should work. Tested with:

• Chrome v65
• Firefox v58
• Safari v11
• Opera v51

No idea about IE/Edge.

Dependencies

Package	Notes	Licence
FileSaver.js	-	FileSaver.js license
jszip	-	MIT
FabricJS	Used instead of some legacy unidentified library from original repository	FabricJS license

Feature

• Basic Pascal VOC and COCO format support.
• Works in your browser on any platform.
• Complete YOLO format support.
• No need for image upload - everything is done locally!
• Zooming and panning images with guidelines for precise bboxing.
• Fast navigation for quick bboxing.
• Auto save in memory in case of accidental refreshes and crashes.
• Ability to crop your bboxes and save the resulting images.
• Information on both image and current bbox.

Caveats

• Loading many and or big images might take a while. This is because tool needs to figure out image dimensions.
• Cropping many items might crash your browser. This and above will be fixed at some point.