Drone-Person Tracking in Uniform Appearance Crowd: A New Dataset

Area of Science:

• Computer Vision
• Artificial Intelligence
• Robotics

Background:

• Drone-person tracking in uniform appearance crowds presents significant challenges.
• Distinguishing individuals with similar attire and handling multi-scale variations are key difficulties.

Purpose of the Study:

• To introduce a novel dataset, D-PTUAC (Drone-Person Tracking in Uniform Appearance Crowd), to address these tracking challenges.
• To facilitate the development and evaluation of advanced drone-person tracking algorithms.

Main Methods:

• The D-PTUAC dataset consists of 138 sequences with over 121,000 frames.
• Each frame is manually annotated with bounding boxes and 18 challenging attributes covering diverse viewpoints and scene complexities.
• Extensive experiments were performed using 44 state-of-the-art (SOTA) trackers.

Main Results:

• The dataset's comprehensive annotations enable performance analysis across various challenging attributes.
• Experiments revealed a performance gap for existing visual object trackers on this dataset compared to others.
• This highlights the limitations of current trackers in aerial tracking scenarios.

Conclusions:

• The D-PTUAC dataset is essential for advancing drone-person tracking research in uniform appearance crowds.
• There is a clear need for dedicated end-to-end aerial visual object trackers that consider the unique properties of aerial environments.