-
Notifications
You must be signed in to change notification settings - Fork 12
User Guide
👋 Welcome to the user guide of using Imannotate.
We'll cover only the usage of the web application of Imannotate here. You can find more information about installation and contribution on other parts of this Wiki.
Just after the instantiation, you'll have to go to your web app and create the first user, which will be the administrator on Imannotate. So it's vital you do it.
Now you have your first user, you can create your first project. First, you have to log-in. Then, click on Create a new project
. You'll have to input a project name, description and the tags to use for annotation. Then you select the input source for your pictures.
ℹ️ So far, we only support Amazon S3 for production. Qwant is for demo purpose only.
Additionally, you can already add member which will be able to annotate pictures right now and upload a picture to represent your project (like a logo or something).
Hit save and go back to the Dashboard by clicking Home
to see your project.
In order to start annotation, click Annotate on the project card you want to annotate. The Annotation interface is loading and the first image is requested from the queue.
ℹ️ We are using a queueing system to ensure all the pictures are annotated and so far annotated just once.
When the picture is loaded, start clicking then drag-drop to create a box. Then select a tag to annotate that box. You can repeat to have all your tag you can see in the picture annotated. If a subject is not visible, that's fine. Then click the Next button to save and get a new picture from the queue.
🤔What if you don't find anything to tag? You can use the Nothing!
button and will have to confirm this picture contain no elements you are asked to annotate.
🤔What if I'm not sure? You can use the Pass
button. It will discard this picture from your assignment and return it to the queue so another user will be able to get it and annotate it.
Nothing!
will tell the model later this picture contains nothing the model should look for, while Pass
has no effect on the model and the quality of its training.
Mady by Smile with ❤️ in Europe 🇪🇺