Video Labeling Editor

Overview of the Video Labeling Editor in Ango Hub

Ango Hub provides a labeling editor with which video files can be annotated. The video editor will be opened when a user opens an asset with the following file extensions: .mp4, .mov, .webm, .ogg, .mkv.

This article will exclusively go over Ango Hub’s video labeling interface. Features common to all labeling editors are instead explained here.

Overview of the Video Labeling Editor in Ango Hub

Playback Bar

The back and forward arrows allow you to move backwards and forwards one frame at a time. The Play button starts and stops playback of the video file. The slider allows you to move between frames by clicking on the playhead and dragging it to find the frame you need.

When dragging with the slider, the frame selected will only be loaded when you release the left mouse button.

The volume slider () allows you to change the volume of the audio when playing the file back.

The slider allows you to change the playback speed, and the slider allows you to change the zoom on the timeline view.

To the left of the back/forward arrows, there is a three-dot menu. The menu options are:

Retry Metadata Extraction: Ango Hub automatically attempts to extract the video's metadata when the video is first uploaded. Metadata includes the video's dimensions, frame rate, and total frame count. If you believe Hub got the frame rate or number of frames wrong, you may force it to retry fetching the metadata from the video, which often solves the issue. After metadata extraction, you will be brought to Frame 1.

Buffering

Ango Hub does not load the entire video in memory right from the start, as that would be computationally expensive, delay loading, and would use unnecessary memory. Instead, Hub loads it chunk by chunk, buffering it.

You can see how much of the video has been downloaded (buffered) by looking at the playback bar as below:

Timeline View

The timeline view allows you to see the annotations throughout the video in a visual way.

When you create a new annotation, a new row will be added to the timeline view. You can click on the row to select its annotation, or click on the annotation to select its row:

Objects belonging to the same class are automatically grouped and collapsed. To see each object, click on the arrow:

Adding a keyframe

When you move the object, or change the answer in a frame-specific classification, a keyframe is added to the row, indicated by a dot on the line:

Removing a keyframe

To remove the keyframe, navigate to the frame where the keyframe is located, and click on the row or the annotation of which you'd like to delete the keyframe. A Delete Keyframe button will appear. Click on it to delete the keyframe:

Stopping an object or classification from appearing

To toggle off an object or classification, disable the toggle present on the object's row. The object will disappear, and the line in its timeline row will be cut off:

How to Annotate Videos

The following are the labeling tools supported on videos:

From the Tools panel on the left sidebar, select a supported labeling tool. Then, follow the instructions found on each tool's docs page, linked to above.

If no tools are present in the project, only answer the questions in the Classifications panel.

Labeling Properties Specific to Videos

Frame Interpolation

Once you have created an object, move forward some frames, then edit the label to its new position on the video. Ango Hub will automatically interpolate the position of the object in the frames where the label was not manually placed.

Interpolation is currently only available for the Bounding Box, Polygon, Segmentation, and Point labeling tools.

If the object you are following with the label leaves the frame, hover over the label with your mouse cursor and press S or disable the toggle next to the annotation from the Objects panel in the lower left of the screen. The label will be hidden from view. To restore it, simply enable its toggle.

Frame-Specific Classifications

When creating classifications such as radio, dropdown, and others, project managers may choose to make those classifications general (e.g., one response per video) or frame-specific (e.g. one response per frame.)

Here's an example on how to classify single frames using frame-specific classifications.

For this example, let's assume I have a video with 506 frames and a radio classification class with the answers "First", "Second", and "Third," and I wish to:

annotate frames 1 to 10 with "First"
frames 11 to 20 with "Second", and
frames 50 to 60 with "Third."

Here's how you would go about doing so:

Navigate to frame 1 and click on First. This will enable the "First" answer for all following frames until we tell Hub otherwise.
Navigate to frame 11 and click on Second. This will switch the annotation to Second.
Navigate to frame 21 and disable the toggle next to the classification. This will make it so that from frame 21 onwards included, the classification will not be answered.
Navigate to frame 50 and click on Third.
Navigate to frame 61 and disable the toggle next to the classification.