GUIDE.md

简体中文 | English

Concept

Concept	Description
Task	Task is created to label a dataset
label	Classification marks to be added to an annotation, such as cat, dog, pedestrian and vehicle
Annotation	Objects generated after a marking operation, such as a rectangular box or a point
Attribute	Further description of a label, such as adding the attribute "vehicle occlusion rate is 20%" after marking an object as a vehicle
Result	Annotation+Label+Attribute, a complete annotation record
Template	The configuration files (XML, JSON, Yaml) that can be referenced when configuring labeling tasks

Register && log in

1. Create an account with email and password
2. Log in to LabelU

Create a new task

1. Click "New Task"
2. Enter the task name, task description, and task hints, among which the task hints will provide hints and help for the labelers during the labeling process
3. Click "Next" to save the basic configuration information and enter the "Data Import" page

Data import

Currently, data import only supports local data import, and supports file or folder import in two ways.

1. Click Upload File or Upload Folder to upload local data
2. Click "Next" to save and enter the label configuration
3. If you want to import data again, you can click "Data Import" on the task home page to enter the data import page to import, and the imported data can be viewed on the task home page

Task configuration

After completing the basic configuration and data import configuration, perform labeling configuration, mainly according to the task scenario to edit the configuration of tools and labels, and configure some other parameters. Currently supports Yaml and visualization two configuration methods, the two methods support linkage, and can preview the home page through the right page

Yaml configuration (to be perfected)

1. Select a template according to the task requirements
2. Customize and modify the Yaml configuration, including tool configuration modification, label configuration modification, other configuration modification, etc.
3. Click Finish

Visual configuration

In addition to using the Yaml configuration, you can also choose to use the visual method to configure the annotations. The configuration process is as follows:

1. Select the annotation type, currently only supports pictures, and will support audio and video, point cloud, text and other types in the future
2. Select the annotation tool, which supports simultaneous configuration of multiple tools (drawing box, polygon, punctuation, marking, classification, description)
3. The configuration required for each tool includes tool properties and label configuration

Configuration instructions for each tool:

Tool	Instruction
rectangle	Minimum size: the minimum width (W) and height (H) of the pull-up box, label configuration: label object classification, including Chinese and English
polygon	Line type: including straight lines and curves, number of closing points: including the minimum number of closing points and the maximum number of closing points, edge adsorption: automatically fit the edge of the object after opening, label configuration: label object classification, including Chinese and English
point	Upper limit of points: the upper limit of specified points, exceeding the limit cannot be drawn, label configuration: label object classification, including Chinese and English
line	Line type: including straight lines and curves, number of closing points: including the minimum number of closing points and the maximum number of closing points, edge adsorption: automatically fit the edge of the object after opening, label configuration: label object classification, including Chinese and English
text	Text list: include text description name in Chinese and English, text setting: include maximum number of words and default text
classification	Category: Category description in Chinese and English, Options: Category specific options in Chinese and English

4. Choose whether to support out-of-target annotations, and configure attributes (currently only text types are supported)
5. Set common labels (when multiple tools share a label system, configuration can be simplified by setting common labels)
6. Preview to view the effect
7. Click Save to enter the task home page

Start labeling

Start labeling tasks according to the task configuration. Each functional area is introduced as follows: Toolbar: perform operations such as tool selection, tool style switching, undo and redo, display order, etc. Label bar: After selecting the tool, click to select the label to label Marking result bar: View and edit the marked results Annotation subject: including drawing area, picture operation

Labeling process description:

1. Determine whether the task is invalid. If it is invalid, click Skip to enter the next task If it is valid, when there are drawing tasks (object detection, semantic segmentation, line labeling, point labeling), the tools and configurations of the drawing tasks are consistent
2. Select Tool
3. Select tab
4. Mark the drawing
5. The attribute information pop-up window will pop up, edit the attribute information, if there is no attribute information, you can leave it blank, click anywhere outside the pop-up window to close.
6. Click the tag result in the tag result management bar on the right to select the corresponding mark in the picture, and switch the tool selected in the toolbar to the tag result tool. Click the [Edit] [Show/Hide] [Delete] button of the tag result to manage it. If valid, when there are label descriptions and classification tasks
7. Fill in the description and classification results in the result management column on the right
8. Select [Next] to enter the next task
9. Repeat 1~8 until the marking is completed

For specific labeling operations of each tool, see:

Image Labeling

Export results

After the annotation is completed, the annotation result file can be exported in the form of Json. The annotation format is described as follows:

LabelU annotation format