OpenTraj: Assessing Prediction Complexity in Human Trajectories Datasets

Javad Amirian (1) , Bingqing Zhang (2) , Francisco Valente Castro (3) , Juan José Baldelomar (3) , Jean-Bernard Hayet (3) , Julien Pettré (1)

1 RAINBOW - Sensor-based and interactive robotics
2 UCL - University College of London [London]
3 CIMAT - Centro de Investigación en Matemáticas

Human Trajectory Prediction (HTP) has gained much momentum in the last years and many solutions have been proposed to solve it. Proper benchmarking being a key issue for comparing methods, this paper addresses the question of evaluating how complex is a given dataset with respect to the prediction problem. For assessing a dataset complexity, we define a series of indicators around three concepts: Trajectory predictability; Trajectory regularity; Context complexity. We compare the most common datasets used in HTP in the light of these indicators and discuss what this may imply on benchmarking of HTP algorithms. Our source code is released on Github.

Mots clés: Human trajectory prediction; trajectory dataset; motion prediction; trajectory forecasting; dataset assessment; benchmarking

Domaines: Robotique [cs.RO]

Javad Amirian, Bingqing Zhang, Francisco Valente Castro, Juan José Baldelomar, Jean-Bernard Hayet, et al.. OpenTraj: Assessing Prediction Complexity in Human Trajectories Datasets. ACCV 2020 - 15th Asian Conference on Computer Vision, Nov 2020, Kyoto / Virtual, Japan. pp.1-17. ⟨hal-03156511⟩