WayveScenes101: A Dataset and Benchmark for Novel View Synthesis in Autonomous Driving

TL;DR

The paper addresses the need for realistic, dynamic driving scenes to benchmark novel view synthesis in autonomous driving. It introduces WayveScenes101, a 101-scene, 101,000-image dataset captured with a five-camera rig, COLMAP poses, and rich per-scene metadata to stress geometry, texture, and occlusion. An off-axis held-out front-forward camera evaluation protocol is proposed to quantify generalization across viewpoints. The work provides open access to the dataset and code, facilitating reproducible benchmarking and advancing research on synthetic-data usage for driving-model evaluation and training.

Abstract

We present WayveScenes101, a dataset designed to help the community advance the state of the art in novel view synthesis that focuses on challenging driving scenes containing many dynamic and deformable elements with changing geometry and texture. The dataset comprises 101 driving scenes across a wide range of environmental conditions and driving scenarios. The dataset is designed for benchmarking reconstructions on in-the-wild driving scenes, with many inherent challenges for scene reconstruction methods including image glare, rapid exposure changes, and highly dynamic scenes with significant occlusion. Along with the raw images, we include COLMAP-derived camera poses in standard data formats. We propose an evaluation protocol for evaluating models on held-out camera views that are off-axis from the training views, specifically testing the generalisation capabilities of methods. Finally, we provide detailed metadata for all scenes, including weather, time of day, and traffic conditions, to allow for a detailed model performance breakdown across scene characteristics. Dataset and code are available at https://github.com/wayveai/wayve_scenes.

Figures (9)

• Figure 1: Overview of the 101 scenes in our dataset.
• Figure 2: scene_024
• Figure 3: scene_032
• Figure 4: scene_071
• Figure 5: scene_100