Object tracking

Object trackers use information from object detectors to track the position and velocity of unique objects in a scene.

API reference

This module contains generic components for tracking algorithms.

class videoanalytics.pipeline.sinks.object_tracking.TrackedObjectsAnnotator(name, context)

Annotates the tracked objects in a frame displaying a bounding box around each identified object with its numeric Id.

This component READS the following entries in the global context:

Variable name

Description

TRACKED_OBJS

Output of an object tracker.

FRAME

Numpy array representing the frame.

This component WRITES the following entries in the global context:

Variable name

Description

FRAME

Numpy array representing the frame.
Parameters

  • name (str) – the component unique name.

  • context (dict) – The global context.

process()

This method is called for each active component in the pipeline.

setup()

This method is called after all components from the pipeline are instanced.

shutdown()

This method is called after the process finished.

class videoanalytics.pipeline.sinks.object_tracking.TrackedObjectsCSVWriter(name, context, filename)

Writes the trackings to a CSV file.

This component READS the following entries in the global context:

Variable name

Description

TRACKED_OBJS

Output of an object tracker.

FRAME

Numpy array representing the frame.
Parameters

  • name (str) – the component unique name.

  • context (dict) – The global context.

  • filename (str) – CSV output file.

process()

This method is called for each active component in the pipeline.

setup()

This method is called after all components from the pipeline are instanced.

shutdown()

This method is called after the process finished.

SORT

Implementation of SORT tracking algorithm.

Note

The code included in this module is only a wrapper class for the original paper’s code with minor changes to adjust documentation format or minor refactoring.

class videoanalytics.pipeline.sinks.object_tracking.sort.SORT(name, context, max_age=1, min_hits=3, iou_threshold=0.3)

Wrapper class for the SORT algorithm.

This component READS the following entries in the global context:

Variable name

Description

DETECTIONS

Output of an object detection model.

START_FRAME

Initial frame index.

This component WRITES the following entries in the global context:

Variable name

Description

TRACKED_OBJS

Object trackings.
Parameters

  • name (str) – the component unique name.

  • context (dict) – The global context.

  • max_age (int) – age limit for a tracklet without being updated before removing it.

  • min_hits (int) – minimum hits required to start a new track.

  • iou_threshold (float) – minimum IoU to consider overlapping boxes.

process()

This method is called for each active component in the pipeline.

setup()

This method is called after all components from the pipeline are instanced.

shutdown()

This method is called after the process finished.

class videoanalytics.pipeline.sinks.object_tracking.sort.Sort(max_age=1, min_hits=3, iou_threshold=0.3)

SORT algorithm implementation.

Parameters

  • name (str) – the component unique name.

  • context (dict) – The global context.

  • max_age (int) – age limit for a tracklet without being updated before removing it.

  • min_hits (int) – minimum hits required to start a new track.

  • iou_threshold (float) – minimum IoU to consider overlapping boxes.

__init__(max_age=1, min_hits=3, iou_threshold=0.3)

Sets key parameters for SORT

update(dets=array([], shape=(0, 5), dtype=float64))

Updates the internal state of the tracked objects. This method must be called once for each frame even with empty detections (use np.empty((0, 5)) for frames without detections).

Parameters

  • - a numpy array of detections in the format [[x1 (dets) –

  • y1 –

  • x2 –

  • y2 –

  • score] –

  • [x1 –

  • y1 –

  • x2 –

  • y2 –

  • score] –

  • ...] –

Returns

A similar array, where the last column is the object ID.

Note: The number of objects returned may differ from the number of detections provided.

videoanalytics.pipeline.sinks.object_tracking.sort.associate_detections_to_trackers(detections, trackers, iou_threshold=0.3)

Assigns detections to tracked object (both represented as bounding boxes) Returns 3 lists of matches, unmatched_detections and unmatched_trackers

DeepSORT

Implementation of DeepSORT tracking algorithm.

Note

The code included in this module is only a wrapper class for the original paper’s code with minor changes to adjust documentation format or minor refactoring.

class videoanalytics.pipeline.sinks.object_tracking.deepsort.DeepSORT(name, context, model_filename, max_cosine_distance=0.4, nn_budget=None, max_iou_distance=0.7, max_age=60, n_init=3)

Wrapper class for the DeepSORT algorithm.

This component READS the following entries in the global context:

Variable name

Description

DETECTIONS

Output of an object detection model.

START_FRAME

Initial frame index.

This component WRITES the following entries in the global context:

Variable name

Description

TRACKED_OBJS

Object trackings.
Parameters

  • name (str) – the component unique name.

  • context (dict) – The global context.

  • model_filename (str) – filename for the reidentification model.

process()

This method is called for each active component in the pipeline.

setup()

This method is called after all components from the pipeline are instanced.

shutdown()

This method is called after the process finished.