WayveScenes101 is a high-resolution image dataset for autonomous driving applications, mainly used for novel view synthesis and scene reconstruction research.
It was released by Wayve Technologies Ltd. and contains high-resolution images and camera position information of 101 different driving scenarios, each lasting 20 seconds. These scenarios cover a variety of different locations, traffic conditions, and environmental conditions, such as different weather, road conditions, and time periods, and are a valuable resource for researching autonomous driving technology.
High-resolution images: The dataset contains 101,000 high-resolution images, each scene recorded by 5 time-synchronized cameras at 10 frames per second.
Diverse scenarios: covers a variety of weather conditions (sunny, cloudy, rainy, foggy), different types of roads (highways, city roads, residential areas, rural roads), and different times (daytime, low sun, night).
Dynamic elements: The scene contains dynamic vehicles, pedestrians, bicycles, animals, etc.
Detailed metadata: Metadata for each scene is provided, including camera calibration data and binary masks for data anonymization.
Image data
High-resolution images
101,000 images in total, distributed across 101 different driving scenarios, each recorded using 5 time-synchronized cameras at 10 frames per second.
Scene diversity
Weather conditions: including sunny, cloudy, rainy and foggy days, etc.
Road conditions: including highways, urban roads, residential areas, rural roads and road construction, etc.
Time period: different time periods such as daytime, low sun and night.
Dynamic elements: dynamic objects such as vehicles, pedestrians, bicycles and animals.
Dynamic lighting: such as traffic lights, exposure changes, lens flares and reflections.
Metadata of WayveScenes101
Camera information
Camera calibration data: includes extrinsic and intrinsic camera calibration parameters, provided in the COLMAP file.
Camera position data: includes the position information of the camera in each scene.
Camera type: 5 time-synchronized cameras: front front, left front, right front, left rear and right rear.
Camera parameters: shutter type (rolling shutter), distortion model (fisheye distortion), resolution (1920 x 1080 pixels), temporal resolution (10 Hz).
Scene attributes
Detailed attribute labels are provided for each scene for fine-grained model evaluation, such as nighttime scenes or rainy scenes.
Dataset File of WayveScenes101
Binary mask
used to mark the annotation location in the image, mainly used to blur license plates and faces for data anonymization, and also marks the location of the ego-car in the camera image.
Tutorials and Usage Guides
Provides detailed tutorial files, for example, tutorial/dataset_usage.ipynb,to guide users on how to view and use the dataset.
Download and Installation Guide of WayveScenes101
Data download: The dataset can be downloaded through Google Drive, and you can choose to download all or part of the scenes.
Environment Setup: Provides a guide to setting up a Conda-based environment, including how to clone a repository, create and activate an environment, and install dependent libraries.
