Object Class Recognition
Object class recognition finds objects that belong to pre-defined classes, when they appear in media. For example, if you are processing video captured by a CCTV camera, you could train Media Server to return the locations of pedestrians, cyclists, and cars that appear in the video.
Configuration Parameter | Description |
---|---|
ClassFilters | A comma-separated list of identifiers that specifies the object classes to recognize. |
DetectionThreshold | The minimum confidence score necessary for Media Server to output a result. |
FieldOfView | The horizontal angle of view of the camera, in degrees. |
FullAnalysisRate | Specifies how frequently to recognize new objects, so that you can optimize performance for your hardware. |
Input | The track to analyze. |
MaxImageSize | The maximum image size for analysis. |
MaxInputQueueLength | Can be used to place a limit on latency. |
MaxPerspectiveError | The maximum percentage difference, between the size of a recognized object and known dimensions for its class. |
MinPerspectiveData | The minimum number of data records that must be generated before attempting to calculate the perspective. |
NumParallel | The maximum number of video frames to analyze simultaneously. |
Recognizer | The recognizer to use for recognizing objects. |
Region | The region of the frame to restrict processing to. |
SampleInterval | The interval at which frames are selected to be analyzed. |
SegmentDuration | The maximum duration of a record in the SegmentedResult or SegmentedResultWithSource track. |
SurveillanceRecognizer | The recognizer to use (to run object class recognition with a surveillance license). |
SyncDatabase | Specifies whether to synchronize with the training database before starting analysis. |
Type | The analysis engine to use. Set this parameter to ObjectClassRecognition . |
WorldDimensions | A list of dimensions that describe the actual size of an object class. |
WorldDimensionsMetadata | The name of a metadata field that describes the actual size of an object class. |
Output Tracks
The following table describes the tracks that are generated by this engine. The Output column indicates whether the information contained in the track is included by default in the output created by an output task (when you don't set the Input
parameter for the output task).
Output track | Description | Output |
---|---|---|
Data
|
Contains one record, describing the analysis results, per recognized object, per video frame. | No |
DataWithSource
|
The same as the |
No |
Result
|
Contains one record, describing the analysis results, for each recognized object. When the same object appears in many consecutive frames, Media Server produces a single result. | Yes |
ResultWithSource
|
The same as the |
No |
SegmentedResult
|
(Available only when processing video). The same as the |
Yes |
SegmentedResultWithSource
|
(Available only when processing video). The same as the |
No |
Start
|
The same as the |
No |
End
|
The same as the |
No |
WorldData
|
These tracks are similar to the Data , Result , and SegmentedResult tracks, respectively. The difference is that each record also contains the approximate position of the object in real-world 3D coordinates. These tracks are available only when you set the Projection parameter or configure Media Server to calculate a perspective projection matrix. For more information about using perspective in media analysis, refer to the Media Server Administration Guide. |
No |
WorldResult
|
No | |
WorldSegmentedResult
|
No |
For more information see Object Class Recognition Results or use the action GetExampleRecord.