Based on large storage and computing servers, processing the 73.25 TB dataset to obtain a final SinoLC-1 land-cover product covering the entire land surface of China, ~9,600,000 km 2, took about 10 months. This framework resolved the label noise stemming from a resolution mismatch between images and labels by combining a resolution-preserving backbone, a weakly supervised module, and a self-supervised loss function, to refine the VHR land-cover results automatically without any manual annotation requirement. These training labels and 1-meter resolution images derived from Google Earth were used to train the proposed framework. Reliable training labels were generated by combining three 10-meter GLC products and OSM data. To fill this gap, the first 1-meter resolution national-scale land-cover map of China, SinoLC-1, was established using a deep learning-based framework and open-access data including global land-cover (GLC) products, open street map (OSM), and Google Earth imagery. Constructing a very-high-resolution (VHR) land-cover dataset for China with national coverage, however, is a non-trivial task and thus, an active area of research impeded by the challenges of image acquisition, manual annotation, and computational complexity. In China, the demand for a more precise perception of the national land surface has become most urgent given the pace of development and urbanization.

Maps earth