Alibaba DAMO Academy pioneers remote sensing AI large-scale model, one-click identification of all things on the ground, empowering Earth science exploration.

Alibaba's DAMO Academy recently launched the industry's first remote sensing AI model, which can quickly identify various objects on the earth's surface such as farmland, crops, and buildings, providing powerful intelligent support for remote sensing applications. This model is now available for use on the AI Earth platform. Remote sensing technology has a wide range of applications in national economy and people's livelihood, such as urban operations, farmland protection, and emergency relief. Remote sensing AI utilizes deep learning, computer vision, and other technologies to analyze massive remote sensing satellite image data in a more in-depth and accurate manner. For example, by combining satellite photos with historical weather conditions, it can predict the growth status of crops in a specific farmland, making agricultural production more scientific and efficient. Traditional remote sensing models face challenges such as massive data scale, complex object classification, low recognition accuracy, and poor generalization. DAMO Academy's remote sensing AI model, called AIE-SEG, solves these problems by achieving unified image segmentation tasks in the field of remote sensing. With this model, "zero-sample extraction" of nearly a hundred types of remote sensing objects can be achieved rapidly and accurately. Furthermore, it can automatically optimize recognition results based on user interactive feedback. In specific scenarios, this model can improve recognition accuracy by 25% to 30% compared to traditional models. The model supports multi-modal interaction. For example, when the user inputs "extract farmland in the image," it will automatically recognize the selected target. The model provides an "out-of-the-box" API calling service, allowing users to customize different remote sensing AI interpretation functions according to their needs, such as water extraction, farmland change monitoring, and photovoltaic identification. In collaboration with DAMO Academy, the Shandong Institute of Land Surveying and Mapping used this model in the monitoring research of winter wheat growth in the entire province of Shandong, achieving an accuracy rate of over 90% and effectively improving the efficiency of remote sensing interpretation for winter wheat. The National Institute of Natural Disaster Prevention and Control utilized this model for landslide and collapsed building identification. In the testing of historical remote sensing images in disaster-prone areas, extracting such disaster information only took a few minutes, which is dozens of times more efficient than manual identification. Luo Hao, the head of the AI Earth algorithm at DAMO Academy's Visual Technology Laboratory, stated that multi-modal remote sensing is the necessary path for humans to better understand the Earth. DAMO Academy will continue to advance the research of remote sensing AI models to assist in the exploration and application of Earth science. AI Earth is a one-stop Earth science cloud platform launched by DAMO Academy in 2022, providing cloud computing and analysis services for multi-source observation data. It has established partnerships with over 50 domestic universities and its related technologies have been applied in institutions such as the Ministry of Water Resources, the National Meteorological Center, and the Ministry of Ecology and Environment.