Video Labelling Overviews
BlueEye Data is also specialized in making annotations for videos with highest accuracy frame by frame.
Supported annotation type:
- Bounding box
- Transcription
- Temporal
To start a data labelling for images project, simply follow these three steps:
Step 1: Click on Project that you are invited to see project info
Step 2: Go to Dataset page by clicking on Dataset menu
This page will display the projects that are assigned to you
Step 3: Go to Labeling tool and start labelling
There are 3 ways to Start Labeling
- Way 1: Click on Start Labeling button on Dataset page
If you are a Reviewer/ Project Manager/ Admin, you can see it as **Start Review** instead. In this case, the system will automatically open the first file on the first dataset on the Labeling tool.
- Way 2: Click on Start Labeling button on a specific dataset page
You can see the Start Labeling button after navigating to one of the datasets listed above.
If you are a Reviewer/ Project Manager/ Admin, you can see it as **Start Review** instead. In this case, the system will automatically open the first file on the first dataset on the Labeling tool.
- Way 3: Click on file name of the file that you want to label on a specific dataset page
In this case, the system will open exactly this selected file on the Labeling tool.
Video Labelling Tool Interface Overviews
This tool is used by Labelers and Reviewers, after starting a project and access labelling tool, this is the interface they will see
It can be broken down into 13 sections:
1. Class list: show all classes needed to label in a project
You can tap on the class field to show all classes created by Project Manager.
When labeling, users will choose the class that needs to be labeled and then choose the corresponding tool.
2. Additional controls: guide link and hotkeys
3. Instruments panel: tools used to create annotations
BlueEye Data maximizes support for Labelers by only showing the supported tools needed for labeling in the Tool panel after users choose appropriate classes. This feature helps users avoid confusion while making annotations.
- Move tool: Allow users to move the main frame on the Editor to see the area that is needed to annotate.
- Bounding box tool: Allow users to draw a bounding box around the object that is needed to annotate.
- Start-stop time tool: Allow users to track start & end time of one action on media.
- Landmark-point tool: Allow users to label for facial landmarks.
- Polygon tool: Allow users to draw a polygon for the object that is needed to annotate.
- Polyline tool: Allow users to draw lines for an object.
- Transcription tool: Allow users to create captions for video frame by frame.
- Cuboid tool: Allow users to draw a cuboid for an object.
- Smart tool segmentation: Allow users to segment the object by modeling auto-generation.
- Smart tool object tracking: Help users label the object automatically after one bounding box is created manually.
Hover on each tool icon on the toolbar will display the name of the tools and hotkeys to choose it.
4. Source list: current video and all videos on dataset
Labeling tool only supports users working on separate datasets. So users can see all assigned sources which are on the current dataset.
Users can preview the media on the list by hovering on it.
Users can track issues on each media with the flag icon attached on each video. This symbol informs the number of open issues/total issues on that media. It helps users know and resolve issues quickly.
5. Submit button: Save and commit your work on current media
6. Main screen: current media and annotation
This is the main working area of users. The purpose of this screen is:
- Display media content
- Tools for drawing bounding box, polygon, etc.
- Display annotation and transcription after users have created
7. Timeline panel: view the bounding box instances throughout the media.
This section allows users to oversee all annotations and transcription in a media.
Users can use the pointer to locate the position of transcription or each annotation that is marked on each bar. Users can also tap on each block on the timeline to find the position of them on the media.
8. Media controller: control media frame by frame
The system allows users to control a media frame by frame. Users can play media normally or the next video by each frame for annotation work.
1. Playback setting button: Allow users to change media speed
2 & 3. Jump backward and forward for 1 second button: Quickly jump to 1 second forward or backward frame from the current frame.
4 & 5. Jump backward and forward for 1 frame button: MOve to the next or previous frame from the current frame.
6. Play button: Play and pause the media.
7. Progress bar: Display progress of playing media in total time.
8. Current time/frame: Display current time and corresponding frame at the pointer. Users can seek the current frame by entering the number of frames on the text box.
9. Common hotkey bar: common hotkeys that use for labeling
10. Spatial annotation panel: list of bounding box instances on the current media
This section displays all annotations that have been created from the Main screen with selected class names.
Each record is one instance of the object that is created by a multi frame bounding box. The selected instance will be synchronously highlighted on the Main screen as well as timeline panel.
11. Transcription panel: list of transcription created on the current media
This section displays all transcriptions that are created by Labelers.
Each record includes transcription and the time range of this text. The selected record will be shown on the Main screen correspondingly.
12. Issue panel: list of issues created on the current media
Support interaction between Reviewers and Labelers when Reviewers assess Labelersβ annotation results. The system helps Reviewers easily create an issue and Labelers track issues on the current media.
13. Window control: hide and show the windows on screen
