Etiketai
etiketai.vercel.app
- About Etiketai
- Why the need of this project?
- In future releases
- Getting started
- Hotkeys
- Contribute
- License
About Etiketai
Etiketai is an online tool designed to label images, useful for training AI models.
It's totally client-side, there is no request to any server. It is implemented with Next.js + Preact.
Annotations are saved as XML files in PASCAL VOC format, the format used by ImageNet. Besides, it also supports YOLO format.
It's strongly inspired by labelImg.
Why the need of this project?
Although we think that labelImg is a good alternative, there are certain root problems that we are trying to solve with this project:
- Available in all devices. Being a web application makes it accessible from any device, even tablets and mobiles.
- No instalation required. It speeds up the start, as it does not require installation and has no dependencies on your operating system. Only the browser.
- Automatic updates. You will always have the latest version available.
- Security. No file on your system is directly manipulated. Files are imported/saved using the security layer of your browser.
- Beginner-friendly. We want it to be an easy-to-use process without losing flexibility. To start, you only need to open a browser with any device.
I wrote a short article in my blog explaining a little more in detail why: https://aralroca.com/blog/etiketai
In future releases
- Improve tablet / mobile experience.
- Possibility to train directly your labeled images with the same app and also to save the generated model.
- Offline support.
Getting started
Etiketai doesn't require any installation
- Open the web app https://etiketai.vercel.app/
- Click 'Open'. Here you can import 3 file types: Images (.png, .jpg, .gif, .svg...), XML (PascalVOC labels) and TXT (YOLO labels). Surely the first time you are only interested in importing images, since you will not have any label saved yet. Important: if you want to import already saved labels, you must import it together with the images at the same time. In the case of YOLO, you also must import the
classes.txtfile. - Click and release left mouse to select a region to annotate the rect box. Then, you can name that label or select one of the already used names.
- Update rect boxes. Once you already labeled a part of the image, you can resize it by clicking on a box corner, move the box using drag & drop, duplicate it (Ctrl + d) or even remove it (⌫ Delete).
- Move through images. Once you finish annotating all the labels of an image, you can press to the
NextandPrevbutton to move through images and start anottating another image. You can also use the arrows → and ←. - Save your changes. You can save the labels of the current file or the labels of all files. At this point, you can decide in which format you want to save them (XML for ImageNet or TXT for YOLO). In the case of YOLO, a file named
classes.txtis saved too and it defines the list of class names that your YOLO label refers to.
Hotkeys
| Key | Action |
|---|---|
| o | Open/import files |
| s | Save |
| d | Duplicate the selected box |
| del | Remove the selected box |
| → | Next image |
| ← | Prev image |
| + | Zoom in |
| - | Zoom out |
| 0 | Reset zoom |
They are applicable when the input-text to name the label is not focused.
Contribute
We encourage you to send any pull-quest, even as a draft. It is a project that we want to grow according to the needs of the community.
License
Contributors ✨
Thanks goes to these wonderful people (emoji key):
 Aral Roca Gomez |
This project follows the all-contributors specification. Contributions of any kind welcome!