CVAT use (1) Introduction to the interface and labeling function

0. Preface

main content:
- Introduce the interface of CVAT annotation.
- Introduces the annotation-related functions supported by CVAT.
Reference materials: Mainly corresponding to the Interface of the annotation tool in the user manual
There are a lot of details, wait until you use it.

The interface is shown in the following figure (the figure is from the user manual, confirmed in the official Demo , and it is still like this), which is divided into five parts, which are introduced below.
Header
- Mainly include several basic functions of CVAT navigation, as well as current user related settings.
- The basic functions mainly include three parts: Tasks/Models/Analytics, which respectively represent labeling tasks/pre-training models/analysis. Finally, I don't know what it is.
Top panel
- Mainly include picture navigation (select picture), Menu, save, undo/restore, full screen, annotation result statistics, mode switching.
- More and more complicated.
Workspace: Workspace, nothing to introduce is the area where the image is located.
Controls sidebar: Control the sidebar of related commands, including setting the picture size, position, creating shape, editing tracks and other basic functions of labeling.
Objects sidebar: label-related sidebar, including two options (objects and labels), as well as appearance related settings (that is, the display color, transparency, etc. of various labels)

I don’t know how to translate it. “Navigation” seems to be unable to fully express the meaning of the word. It is probably operations such as selecting an image frame and selecting the location of certain areas in the image frame.
Select image frame related
- Button functions include 前/后一帧, 前/后 step 帧, 第一帧/最后一帧,播放
- There is also a progress bar indicating the position of the image frame in all data, and the frame number.
After selecting the frame, there are several basic operations to select the position of the image and zoom
- Move image/choose center position: Use the option of the first image.
- The other two functions are: adapt the image to the screen size/select an area to enlarge.

Shape Mode
- Function: Add annotations to a group of pictures, and modify the existing annotations
- Use: When creating various shapes, there are two modes: shape and track. The shape mode is shape mode.
- More details will be discussed later.
Track Mode
- Function: Used to label the track, the essence is to label the key frames, and the labeling between the key frames is realized by the difference.
- Use: When creating various shapes, select the track mode, just like the above picture, don't put it.
Attribute annotation mode
- Function: It can quickly mark the object's annotation and the corresponding attribute value through the keyboard.
- Turn on: You need to select this mode in the top pannel, pay attention to the upper right of the figure below.

Download annotation results
- The approximate steps are: save your thoughts and select the required structure in the Menu.
Task synchronization:
- I don’t know the specific details of this, maybe it is to synchronize the task to a branch of the git project? Currently not used, no further understanding.
- [External link image transfer failed. The source site may have an anti-leech link mechanism. It is recommended to save the image and upload it directly (img-obXyVSmT-1606707111630)(E:\blogs\images\image106.jpg)]
Vocabulary (tag structure)
- label: Define the label target type
- attribute: The attribute of the object, with options such as single selection and multiple selection.
- track: It is the association of bboxes between multiple frames.
- Annotation: can be understood as a collection of a series of shapes and tracks.