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:

1. Should support object detection and segmentation problems
2. Should work offline on my PC (windows, possible mac)
3. The ability to create databases for Mask R-CNN, U-Net, YoloX models
4. 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

https://github.com/heartexlabs/awesome-data-labeling