It’s a quick note about tools which I’m using for creation datasets.
The goal: I would like to have a tool or tools to create datasets for computer vision models.
My requirements for these tools:
- Should support object detection and segmentation problems
- Should work offline on my PC (windows, possible mac)
- The ability to create databases for Mask R-CNN, U-Net, YoloX models
- Will be great to avoid payments 🙂
Which tools acceptable for me:
- https://github.com/wkentaro/labelme Nice offline tool. Can be installed
conda create --name=labelme python=3
conda activate labelme
pip install labelme
labelme --help
Export to COCO & Masks https://github.com/wkentaro/labelme/tree/main/examples/instance_segmentation
-
https://supervise.ly/ It is online service and not so intuitive to use but has free plan and can be used for quick prototyping. Can export dataset as COCO or masks.
-
https://labelbox.com/ Online, but flexible labelling editor which can be configured for any purposes. Has free plan. Not so easy export procedure.
-
still in search
🙂
Other tools:
A list of awesome data labelling tools