车辆图像字幕数据集

该数据集是全面的车辆图像字幕数据集，用于助力了解道路场景，包含在印度不同道路场景中拍摄的 1000 + 张高分辨率图像，且图像来自印度各个城市和地区，涵盖白天和黑夜场景、不同距离、不同背景及视角；每张图像都附有人类生成并验证的详细标题，标题遵循特定问题格式，从左到右描述道路上的每个物体，包含车辆颜色、挡风玻璃外观等多类信息，单条标题长度达 60-70 字，能提供丰富上下文；同时，数据集还包含 15 + 类印度车辆的边界框注释This is a comprehensive vehicle image captioning dataset designed to facilitate road scene understanding. It contains over 1,000 high-resolution images captured in diverse road scenarios across various cities and regions of India, covering daytime and nighttime conditions, different shooting distances, varied backgrounds and perspectives. Each image is paired with detailed captions generated and verified by humans, which follow a specific question-based format to describe every object on the road from left to right, including multiple types of information such as vehicle color, windshield appearance and more. Each individual caption ranges from 60 to 70 words in length, providing rich contextual details. Additionally, the dataset includes bounding box annotations for more than 15 categories of Indian vehicles.

搜集汇总

背景与挑战

背景概述

该数据集是一个全面的车辆图像字幕数据集，包含1000+张在印度不同道路场景中拍摄的高分辨率图像，涵盖多种环境和视角。每张图像都附有详细的人类生成标题和边界框注释，适用于生成式人工智能、自动驾驶和场景理解等任务。