国家尺度的农业景观理解
计算机视觉与模式识别
2024-11-11 v1 人工智能
计算机与社会
摘要
农业景观十分复杂，尤其是在全球南方地区，那里的农田面积更小，农业实践更加多样化。本文报告了我们在印度研究区域中对农业景观（自然与人工）进行数字化处理的进展。我们使用高分辨率影像和 UNet 风格的分割模型，生成了首个国家级多类全景分割输出。通过这项工作，我们能够在 1.517 亿公顷的范围内识别单个农田，并勾勒出水资源和植被等关键特征。我们分享了该输出如何由我们的团队和下游用户进行验证，包括一些可导向针对性数据驱动决策的示例用例。我们相信该数据集将通过生成基础底图，为农业数字化做出贡献。
引用
@article{arxiv.2411.05359,
title = {Agricultural Landscape Understanding At Country-Scale},
author = {Radhika Dua and Nikita Saxena and Aditi Agarwal and Alex Wilson and Gaurav Singh and Hoang Tran and Ishan Deshpande and Amandeep Kaur and Gaurav Aggarwal and Chandan Nath and Arnab Basu and Vishal Batchu and Sharath Holla and Bindiya Kurle and Olana Missura and Rahul Aggarwal and Shubhika Garg and Nishi Shah and Avneet Singh and Dinesh Tewari and Agata Dondzik and Bharat Adsul and Milind Sohoni and Asim Rama Praveen and Aaryan Dangi and Lisan Kadivar and E Abhishek and Niranjan Sudhansu and Kamlakar Hattekar and Sameer Datar and Musty Krishna Chaithanya and Anumas Ranjith Reddy and Aashish Kumar and Betala Laxmi Tirumala and Alok Talekar},
journal= {arXiv preprint arXiv:2411.05359},
year = {2024}
}
备注
34 pages, 7 tables, 15 figs