D-PTUAC

Surveillance Settings

A drone person tracking dataset should include both static and dynamic drone scenarios to evaluate tracking algorithms in diverse operational conditions. Static scenarios assess algorithms' performance when the drone remains stationary while tracking a person, evaluating their ability to handle challenges such as multi-scale variations. Dynamic scenarios evaluate algorithms' adaptability and accuracy in dynamic environments where the drone is in motion while tracking a person, evaluating their ability to handle challenges such as fast motion. By including both static and moving scenarios, the dataset provides a comprehensive evaluation of tracking algorithms' capabilities and limitations in real-world situations.

Attribute Distribution

Comparison of sequence distrubution in each attribute between training and testing sets

The D-PTUAC dataset was divided into training and testing sets. The training set consisted of 90 videos, while the testing set included 48 videos. The training set contained approximately 78k frames, while the testing set had around 42k frames. The ratios of sequences across all 16 attributes and 12 different settings show similarity between the taining and testing sets which provides evidence of the consistency and coherence of our training/testing split.

Daytime Scenarios

Including both morning and evening scenarios in a drone person tracking dataset is crucial for evaluating tracking algorithms' performance under varying lighting conditions. The differences in shadows, illumination levels, and visibility between morning and evening provide insights into the algorithms' adaptability and robustness in real-world scenarios. Assessing trackers in diverse scenarios helps identify challenges and limitations, facilitating the development of more effective drone-person tracking systems.