frames

In the VisionAI format, a frame serves as a container for dynamic, time-based information, encompassing both object and context data. They are used to represent data that is dynamic and time-based in nature, such as information about the location, velocity, and orientation of objects within a sequence.

Each frame is represented by a series of 12 digital numbers that are arranged in order to indicate the temporal relationship.

Schema

Example

"frames": {
    "000000000000": {
        "frame_properties": {
            "timestamp": "2020-04-11T12:00:01.000+0100",
            "streams": {
    	        "camera1": {
                    "uri": "https://image1.path.url.png",
                    "stream_properties": {
          	            "sync": {"timestamp": "2020-04-11T12:00:07.010+0100"}
                    }
                },
                "lidar1": {
                    "uri": "https://lidar0.path.url.pcd",
                    "stream_properties": {
                        "sync": {"timestamp": "2020-04-11T12:00:10.087+0100"}
                    }
                }
            }
        },
        "contexts": {
            "3456a995-a39c-41f2-a9ab-a4c9be693456": { ... },
            "6789a995-a39c-41f2-a9ab-a4c9be697892": { ... }
        },
        "objects":{
            "2eeda995-a39c-41f2-a9ab-a4c9be695beb": {...},
            "cf66fbb2-30b7-43b9-96db-85f70b3619f9": {...}
        }
    },
    "000000000001": {...}
}

frames {}

name	description	type	required
${FRAME_ID}	The id of frame in this annotation. It is the frame order containing 12 digital numbers. (ex. 000000000000, 000000000001, 000000000002)	object	true
frame_properties	Show the information of this frame.	object	true
contexts	Shows all contexts’ dynamic information in this frame. Refer to frames-contexts.	object	false
objects	Shows all objects’ dynamic information in this frame. Refer to frames-objects.	object	false

frame_properties {}

It describes this frame’s information including the reference time, and corresponding streams.

name	description	type	required
timestamp	A relative or absolute time reference that specifies the time instant this frame corresponds to. Representation of date and time format: "yyyy-MM-ddTHH:mm:ss.FFFZ" （ex. 2021-09-03T11:23:56.000+0100）	time	false
streams	The “streams" information refers to the streams(sensors) that were used to capture data in this particular frame. This information helps to identify the sources of data that are included in that frame.	object	true

streams {}

It describes this frame’s stream information including the corresponding stream name and related data path.

name	description	type	required
${STREAM_NAME}	The name of the stream under this frame. All the streams are matched and described under the visionai - streams.	object	true
uri	The url or file path for this stream data.	object	true
stream_properties	Additional properties of this stream. ex. sync infomation	object	false

stream_properties {}

It contains the additional properties of this stream including the absolute time of the sensor in this frame.

name	description	type	required
sync	Define the frame number of the stream that corresponds to this frame, along with timestamping information.	object	false
timestamp	An absolute time reference that specifies the time instant from the sensor of this frame. Representation of date and time format: "yyyy-MM-ddTHH:mm:ss.FFFZ" （ex. 2021-09-03T11:23:56.000+0100）	time	false

---

Use Case

More infomation on

Use Case

bbox

To describe a bbox dataset with one camera sensor:

bbox

bbox + cuboid (3D)

To describe a dataset with one camera sensor (bbox annotation) and one lidar sensor (cuboid annotation) in the coordinate system of iso8855-1:

bbox + cuboid (3d)

semantic segmentation

To describe a semantic segmentation dataset with one camera sensor:

semantic segmetation

tagging

To describe a dataset with taggings:

tagging