High-Resolution Satellite Imagery Dataset of the Chettinad Region, Southern India Captured Using Google Earth Pro

This dataset comprises high-resolution satellite imagery of the Chettinad region in Tamil Nadu, India, collected using Google Earth Pro. Chettinad is renowned for its cultural heritage, traditional architecture, dryland agriculture, and semi-arid landscapes, making it an ideal case for geospatial research. The dataset captures diverse land cover types such as drylands, vegetation patches, water bodies, and built-up settlements with traditional Chettinad mansions. Designed for academic research in remote sensing, computer vision, and multimodal AI, the dataset supports tasks like semantic segmentation, land-use classification, vision-language modeling (VLM), image captioning, and geospatial question answering. Each image has been manually curated to ensure spatial diversity and visual clarity. All the images are in JPG format. Each image is provided at consistent scale (with visible scale bars ranging from 40m to 100m) and orientation (with north indicators), making it suitable for spatially coherent learning tasks. The dataset emphasizes real-world satellite perspectives without artificial annotations, supporting unsupervised and weakly-supervised learning strategies.

本数据集包含印度泰米尔纳德邦切蒂纳德（Chettinad）区域的高分辨率卫星影像，采集自谷歌地球专业版（Google Earth Pro）。切蒂纳德以其文化遗产、传统建筑、旱地农业与半干旱地貌闻名，是开展地理空间研究的理想案例区域。该数据集涵盖多样的土地覆盖类型，包括旱地、植被斑块、水体以及带有传统切蒂纳德宅邸的建成聚落。本数据集专为遥感、计算机视觉与多模态人工智能（multimodal AI）的学术研究而设计，可支持语义分割、土地利用分类、视觉语言建模（Vision-Language Modeling, VLM）、图像字幕生成以及地理空间问答等任务。所有影像均经过人工甄选，以保障空间多样性与视觉清晰度。所有影像均采用JPG格式存储。每张影像均采用统一的比例尺（可见比例尺尺范围为40米至100米）与朝向（带有北向指示标识），适配需要空间一致性的学习任务。本数据集保留真实的卫星拍摄视角，未添加人工标注，可支持无监督学习与弱监督学习策略的相关研究。