GOSPA and T-GOSPA quasi-metrics for evaluation of multi-object tracking algorithms
Researchers have introduced two new quasi-metrics, GOSPA and T-GOSPA, designed to evaluate the performance of multi-object tracking algorithms. These metrics extend existing GOSPA and T-GOSPA measures by allowing for flexible penalties on missed and false objects, and non-symmetric localization costs. The T-GOSPA quasi-metric additionally incorporates a cost for track switching. Simulations were conducted to assess various Bayesian MOT algorithms using the T-GOSPA quasi-metric. AI
IMPACT Introduces new evaluation metrics for multi-object tracking, potentially improving algorithm development and comparison.