Yolo based object detector estimates uncertainty for better reliability
Ev-YOLO: Uncertainty-Aware Object Detection via a Unified Evidential Formulation
Computer Vision and Pattern Recognition
Summary
Object detectors need to know when they might be wrong to work well in uncertain environments, like self-driving cars. This paper presents a version of YOLOv8 that not only detects objects but also estimates how uncertain it is about both what the object is and where it is. The method uses a shared mathematical framework to give estimates of belief, uncertainty, and probabilities for both classification and location. Tests show it keeps detection quality while giving useful uncertainty estimates that help tell true detections from errors, especially when conditions change.
What this means in practice
- •For autonomous vehicle engineers: Integrate uncertainty-aware object detection to improve safety by distinguishing reliable from unreliable detections in real-world driving scenarios.
- •For surveillance system developers: Deploy detectors that quantify localization confidence to better recognize when object tracking or alerts may be inaccurate due to environmental changes.
Authors
Simon Barbarit-Gaboriau, Hind Laghmara, Rémi Boutteau, Samia Ainouz
Abstract
Reliable uncertainty estimation is essential for deploying object detectors in autonomous systems operating in uncertain environments. Evidential Deep Learning (EDL) provides a principled framework for uncertainty-aware classification by representing network outputs as evidence and interpreting predictions through subjective logic. However, existing evidential object detectors typically combine evidential classification with regression uncertainty models that do not share the same theoretical foundation. In this work, we propose an evidential version of YOLOv8 in which both classification and bounding-box regression are formulated within a common evidential framework. Our approach exploits YOLOv8's distribution-based bounding-box representation, allowing the evidential formulation to be applied not only to classification but also to localisation. As a result, both tasks produce belief, uncertainty, and probability estimates that can be interpreted within the Dempster--Shafer framework. Experiments on KITTI, MUSES, and nuScenes show that the resulting detector remains broadly competitive with standard YOLOv8 in terms of detection accuracy while providing a localisation uncertainty that effectively discriminates between correct and erroneous detections. Moreover, this uncertainty becomes increasingly discriminative under domain shift.