斑马线图像数据集

该数据集由 DataCluster Labs 收集，是一组极具挑战性的数据集，包含来自多个建筑工地的 3000 + 张挖掘机车辆图像；图像从 200 + 个不同地点捕获并经众包获取，由 1000 + 众包贡献者参与，每张图像均经 DataCluster Labs 的计算机视觉专业人员手动审查和验证；其特征包括高清及以上分辨率（1920x1080 及以上）、涵盖白天黑夜等各种照明条件与不同距离、视点，且使用 2020-2022 年的手机拍摄

Collected by DataCluster Labs, this is a highly challenging dataset containing over 3,000 excavator vehicle images sourced from multiple construction sites. The images were captured at more than 200 distinct locations and acquired through crowdsourcing, with participation from over 1,000 crowd contributors. Each image has been manually reviewed and validated by computer vision professionals at DataCluster Labs. The dataset features images with high-definition and higher resolutions (1920x1080 or above), covering various lighting conditions including daytime and nighttime, as well as different distances and viewpoints. All images were captured using mobile phones between 2020 and 2022.

搜集汇总

背景与挑战

背景概述

该数据集包含3000+张高清分辨率的挖掘机车辆图像，涵盖多种照明条件和视角，并提供COCO、YOLO、PASCAL-VOC、Tf-Record四种注释格式文件，适用于生成式人工智能、视觉问答等技术的研发。