Evaluation Metrics for Multiple Object Tracking
Learn different Metrics for evaluating the performance of a Multiple Object Tracking Algorithm
Multiple Object Tracking(MOT) is the task of detecting various objects of interest in a video, tracking these detected objects in subsequent frames by assigning them a unique ID, and maintaining these unique IDs as the objects move around in a video in successive frames.
Multiple Object Tracking(MOT) finds its application in video surveillance, robotics, or self-driving vehicles.
To understand the different metrics used to evaluate MOT algorithms, you first need to understand how MOT works.
MOT takes a single continuous video as an input and splits it into discrete frames at a specific frame rate(fps). The output of MOT is
- Detection: what objects are present in each frame
- Localization: where objects are in each frame
- Association: whether objects in different frames belong to the same or different objects
You are performing sports analysis using Multi-object tracking. Would you like to detect every object in the frame precisely or detect the players and their trajectories?
What is more critical for a self-driving car using a Multi-object tracking algorithm is to detect every pedestrian to avoid a collision or correctly associate the objects detected over time?
For video surveillance, is it essential to ensure all objects are detected, and their trajectories tracked accurately?
Read on to find out which MOT evaluation metrics best applies to the above scenarios.
To evaluate the performance of MOT algorithms, measure how well a tracker performs by comparing its predictions to the ground-truth set of tracking results.
Characteristics of MOT Evaluation Metrics
MOT evaluation metrics need to exhibit two significant properties
- MOT evaluation metrics need to address five error types in MOT. These five error types are False negatives(FN), False positives(FP), Fragmentation, Mergers, and Deviation.