Generating realistic road scenes is crucial for advanced driving systems, particularly for training deep learning methods and validation. Numerous efforts aim to create larger and more realistic synthetic datasets using graphics engines or synthetic-to-real domain adaptation algorithms. In the realm of computer-generated images (CGIs), assessing fidelity is challenging and involves both objective and subjective aspects. Our study adopts a comprehensive conceptual framework to quantify the fidelity of RGB images, unlike existing methods that are predominantly application-specific. This is probably due to the data complexity and huge range of possible situations and conditions encountered. In this paper, a set of distinct metrics assessing the level of fidelity of virtual RGB images is proposed. For quantifying image fidelity, we analyze both local and global perspectives of texture and the high-frequency information in images. Our focus is on the statistical characteristics of realistic and synthetic road datasets, using over 28,000 images from at least 10 datasets. Through a thorough examination, we aim to reveal insights into texture patterns and high-frequency components contributing to the objective perception of data realism in road scenes. This study, exploring image fidelity in both virtual and real conditions, takes the perspective of an embedded camera rather than the human eye. The results of this work, including a pioneering set of objective scores applied to real, virtual, and improved virtual data, offer crucial insights and are an asset for the scientific community in quantifying fidelity levels.

With the development of AI-based perception algorithms using cameras, access to large and representative datasets is crucial. For autonomous driving systems, it is essential to use road context datasets covering the entire Operating Design Domain, including various road configurations, driving scenarios, and weather conditions. In this context, it is imperative to propose mechanisms and metrics that allow quantifying the fidelity and the level of representativeness of these simulated datasets, in order to evaluate and validate their usability for training and evaluation stages. In this paper, we propose an objective and multi-modal approach allowing to calculate 4 scores representing several aspects of the synthetic image fidelity. These scores address local, global and statistical texture analysis. In addition, a multi-criteria approach, based on evidence theory, is proposed to merge these scores to obtain a final global score. The result is the generation of a global score along with the uncertainty and conflict quantification. This method has been applied on a large number of real and virtual datasets in different weather conditions (clear, rainy, foggy). The initial results are promising and confirm the interest and the relevance of this method.

Data association in multi-object tracking (MOT) can be viewed as an evidence fusion problem, where heterogeneous cues provide incomplete, uncertain, and sometimes conflicting information for candidate associations. Existing tracking-by-detection (TBD) methods typically compress these cues into a scalar association score before assignment, potentially discarding important structural information in the association evidence. This paper proposes an evidence fusion framework for MOT data association based on belief theory. The framework uses belief masses to represent different forms of association evidence, including support, opposition, residual uncertainty, conflict, and unmatched cases. It further introduces a hierarchical reliability modeling scheme that evaluates cue-level reliability from intra-cue ambiguity and inter-cue disagreement, while estimating association-level reliability for candidate associations. Based on this representation, a two-stage fusion mechanism is developed to combine heterogeneous cues and account for competition among association hypotheses. The proposed framework is designed as a plug-in association module for SORT-like trackers and requires no additional training. Experiments across multiple benchmarks, trackers, and detector backbones show consistent improvements across tracking metrics (e.g., HOTA, AssA, and IDF1), demonstrating the effectiveness of the proposed framework under challenging association conditions.

With the development of AI-based perception algorithms using cameras, access to large and representative datasets is crucial. For autonomous driving systems, it is essential to use road context datasets covering the entire Operating Design Domain (ODD), including various road configurations, driving scenarios, and weather conditions.In this context, it is imperative to propose mechanisms and metrics that allow quantifying the fidelity and the level of representativeness of these simulated datasets, in order to evaluate and validate their usability for training and evaluation stages.In this paper, we propose an objective and multi-modal approach allowing to calculate 4 scores representing several aspects of the synthetic image fidelity. These scores address local, global and statistical texture analysis. In addition, a multi-criteria approach, based on evidence theory, is proposed to merge these scores to obtain a final global score. The result is the generation of a global score along with the uncertainty and conflict quantification. This method has been applied on a large number of real and virtual datasets in different weather conditions (clear, rainy, foggy). The initial results are promising and confirm the interest and the relevance of this method.

Developing and evaluating visual perception systems for autonomous vehicles requires data across diverse and adverse driving conditions, yet collecting and annotating such real-world data is costly and often impractical. To address this challenge, we propose a modular, scenario-driven framework for generating synthetic datasets tailored to the evaluation of visual perception functions. The framework aligns with the operational boundaries and detection–response requirements of automated driving functions and comprises three stages: (1) configuring use-case-driven scenarios, (2) generating sensor data and ground truth via simulation, and (3) post-processing to ensure dataset usability. Designed to be generic and flexible, the framework is instantiated and demonstrated through its integration with specific platforms and tools, namely Pro-SiVIC and RTMaps. We evaluate the generated dataset from two perspectives, image fidelity and perception performance under synthetic weather conditions, in comparison to real-world conditions. Furthermore, we train multiple perception models under different learning paradigms, including baseline, transfer-learning, and mixed-training strategies, to examine the influence of synthetic data on robustness. Experimental results demonstrate not only the high quality of the generated data but also its effectiveness in evaluating visual perception functions, as well as its benefit to model robustness and generalization.