Abstract
Real-time object tracking necessitates a delicate balance between speed andaccuracy, a challenge exacerbated by the computational demands of deep learningmethods. In this paper, we propose Confidence-Triggered Detection (CTD), aninnovative approach that strategically bypasses object detection for framesclosely resembling intermediate states, leveraging tracker confidence scores.CTD not only enhances tracking speed but also preserves accuracy, surpassingexisting tracking algorithms. Through extensive evaluation across varioustracker confidence thresholds, we identify an optimal trade-off betweentracking speed and accuracy, providing crucial insights for parameterfine-tuning and enhancing CTD's practicality in real-world scenarios. Ourexperiments across diverse detection models underscore the robustness andversatility of the CTD framework, demonstrating its potential to enablereal-time tracking in resource-constrained environments.