Skip to content

Latest commit

 

History

History
112 lines (69 loc) · 5.09 KB

admin-guide.md

File metadata and controls

112 lines (69 loc) · 5.09 KB

Pixano

A Guide for Pixano App Configuration and Administration

This guide will help you to configure your annotation project.

The screenshots on this page are all generated using the "data-test" test data set. You can get the exact same results by running Pixano-app with the following command:

node server/server.js ./data-test

Table of Contents


Login and first step

pixano-elements

First authentication is: username: admin password: admin.

After login in as admin, Pixano's dashboard will be displayed.

pixano-elements

You can then create your annotation project in the Tasks tab and update your login or create new users in the Users tab.

Create new tasks and load linked datasets

pixano-elements

Create your first task by clicking on NEW TASK.

pixano-elements

You can now fill the task configuration as follows. When clicking on CREATE TASK Pixano will load all the images in the Data folder and create a task with the name Task name. For each image, an annotation job will be created and served to the annotators.

NB: Your images folder refers to the folder containing the images you want to annotate by a relative path to the workspace folder (in this example, the workspace is ./data-test).

You can define as many tasks as needed. They can refer to different datasets or not. For each defined task, you can:

  • delete them one by one by clicking on REMOVE TASK
  • modify their attributes and saving your changes by clicking on SAVE TASK

A task is defined by:

  • Task name: the name you want to give to this task
  • Data folder: the folder containing the data to be annotated in this task (relative path to the workspace folder)
  • Plugin: the plugin/functionality to be used for this task (choose it from the list). For example, if you want to annotate vehicles by rectangles, choose rectangle. Get more information on plugins in our plugins' guide.
  • Left panel: a form that gives the annotation attributes: for each class defined, you can choose a color and properties of various formats. Define as many classes you need for your task. A default form is provided that gives you all the options available for the selected plugin.
  • Right panel: only used in the case of a smart plugin. Place here the path of the model to be used (or use the default model)

Your datasets

Each annotation task refers to a defined dataset. The same dataset can be used for multiple tasks.

A dataset is a folder containing images (directly or inside a hierarchy of folders).

For example, your data-test folder could have the following structure:

data-test   
│
└───images
    │   xxx.jpg
    │   yyy.jpg
    │
    │   ...

About data format:

  • Make sure your image extensions are either png or jpg.
  • Videos are not directly handled: extract the video frames beforehand. For every plugin taking a sequence as input (e.g. sequence-rectangle, sequence-polygon, tracking, etc), each subfolder containing images will be considered as a sequence.

Create new users

pixano-elements

Define your users and their role: administrator (Admin) or annotator (User); and clic on ADD.

Dashboard: View / Annotate / Validate

Once you defined some tasks, your dashboard will show up similar to this:

pixano-elements

You can select a task to be or being processed (and its linked dataset to be viewed) through the drop list Task. You are now able to navigate in your dataset and view its images and see for each of them its status (annotated, validated, etc). The filtering bar enables you to search for a subpart of your dataset.

As an admin:

  • using the START ANNOTATING button, the annotation work can begin following the previously defined tasks.
  • using the START VALIDATING button, you will be able to validate the data that has been annotated so long.

If you defined users, these users will be able to connect to Pixano and start annotating (see our annotator's guide). For now, only an admin is allowed to validate the annotations.