Object Detection With Probabilistic Guarantees

Florence de Grancey (1) , Jean-Luc Adam (2) , Lucian Alecu (3) , Sébastien Gerchinovitz (4, 5) , Franck Mamalet (4) , David Vigouroux (4)

1 Thales AVS France SAS
2 RENAULT
3 Continental Automotive France [Toulouse]
4 IRT Saint Exupéry - Institut de Recherche Technologique
5 IMT - Institut de Mathématiques de Toulouse UMR5219

Providing reliable uncertainty quantification for complex visual tasks such as object detection is of utmost importance for safety-critical applications such as autonomous driving, tumor detection, etc. Conformal prediction methods offer simple yet practical means to build uncertainty estimations that come with probabilistic guarantees. In this paper we apply such methods to the task of object localization and illustrate our analysis on a pedestrian detection use-case. We highlight both theoretical and practical implications of our analysis.

Mots clés: Object detection; Conformal prediction; Uncertainty quantification

Domaines: Machine Learning [stat.ML]; Intelligence artificielle [cs.AI]

Florence de Grancey, Jean-Luc Adam, Lucian Alecu, Sébastien Gerchinovitz, Franck Mamalet, et al.. Object Detection With Probabilistic Guarantees: a Conformal Prediction Approach. Fifth International Workshop on Artificial Intelligence Safety Engineering (WAISE 2022), Sep 2022, München, Germany. ⟨hal-03769683⟩