A visualization of the CarPatch data: RGB images (left), depth images (center), and semantic segmentation of vehicle components (right).

Abstract

Neural Radiance Fields (NeRFs) have gained widespread recognition as a highly effective technique for representing 3D recon structions of objects and scenes derived from sets of images. Despite their efficiency, NeRF models can pose challenges in certain scenarios such as vehicle inspection, where the lack of sufficient data or the presence of challenging elements (e.g. reflections) strongly impact the accuracy of the reconstruction. To this aim, we introduce CarPatch, a novel syn- thetic benchmark of vehicles. In addition to a set of images annotated with their intrinsic and extrinsic camera parameters, the corresponding depth maps and semantic segmentation masks have been generated for each view. Global and part-based metrics have been defined and used to evaluate, compare, and better characterize some state-of-the-art tech- niques. The dataset will be publicly released to be used as an evaluation guide and as a baseline for future work on this challenging topic.

CarPatch dataset components:

Acknowledgements

We recognize the authorship of the following 3D models:

• Tesla 2018 Model 3 by Ameer Studio
• Smart By Material by Beztao Centro de Diseno
• Ford Raptor-insomnia fully rigged by Eyasu Biyaylgn
• BMW M3 E46 by Marcio Meireles
• Mercedes GLK by VRA
• Volvo S90 Recharge by Lazercar
• Jeep Compass 2022 Lowpoly by Davidson

All the models are licensed under Creative Common Attribution: {http://creativecommons.org/licenses/by/4.0/}. Attribution 4.0 International (CC BY 4.0).