CARLA Dataset - Object Detection
收藏Zenodo2026-06-02 更新2026-05-29 收录
下载链接：
https://zenodo.org/doi/10.5281/zenodo.20071988
下载链接
链接失效反馈官方服务：
资源简介：
The CARLA (Cardilla Recognition and Localization Analyzer) dataset contains drone delivered imagery used for developing computer vision models (object detection problem) in order to study and advance how to imagery and video data can help in detecting weeds (cardilla, Eryngium horridum) in rangelands.
The current repository contains two different types of dataset: a non-tiled version (images are in 1920x1080px) and a tiled version (images are in 640x640px).
Non-tiled version: images have not been processed. They can be considered "raw", obtained directly from the drone video recorded. A total of 900 images with 34,200 labelled instances.
Tiled version: every single frame of the non-tiled version was divided in a grid, as every image needed to be 640x640px with an overlap of 160px between each image. A total of 10,928 images with 87,408 labelled instances.
Both versions have the following data augmentations:
Flip: Horizontal, Vertical
90° Rotate: Clockwise, Counter-Clockwise
Blur: Up to 2.5px
提供机构：
Zenodo创建时间：
2026-05-07
