DAMO Academy releases the industry's first large-scale remote sensing AI model, claiming to be able to identify nearly 100 types of land object classifications

IT House News on October 20, according to the official public account of DAMO Academy, Alibaba DAMO Academy today released the industry’s first remote sensing AI large model (AIE-SEG), claiming to be “the first to achieve this in the field of remote sensing. "The task of image segmentation is unified", "One model realizes the rapid extraction of 'zero samples of everything'", and can identify nearly 100 types of remote sensing land object classifications such as farmland, water areas, buildings, etc., and can also automatically tune based on user interactive feedback Recognition results.

It is reported that remote sensing technology is mainly used in urban planning, farmland protection, emergency disaster relief and other industry applications. With the support of AI, relevant remote sensing technology can analyze satellite capture content and historical meteorological data, thereby assisting urban operations and farmland protection. , emergency disaster relief and other industry applications.

IT House summarizes the characteristics of this large remote sensing model as follows:

• Supports multi-modal interaction. For example, if you enter "Extract farmland from images", the selected target will be automatically recognized

▲ Picture source DAMO Academy DAMO official public account

• Support any surface target recognition and establish a multi-level semantic tag system

▲ Picture source DAMO Academy DAMO official public account

• Supports full element extraction including satellite and drone images

▲ Picture source DAMO Academy DAMO official public account

• Supports interactive result correction. For example, when identifying "paddy fields" and mistakenly extracting "water areas", the deviation can be corrected through manual operation

▲ Picture source DAMO Academy DAMO official public account

• Supports general and multi-category change detection

▲ Picture source DAMO Academy DAMO official public account

Officials stated that in some specific scenarios, compared with traditional remote sensing models, the accuracy of instance extraction can be increased by 25%, and the accuracy of change detection can be increased by 30%.

Damo Academy also claimed that this large remote sensing AI model provides "out-of-the-box" API calling services, and users can customize different remote sensing AI interpretation functions according to their needs, such as "water body extraction", "cultivated land changes" Monitoring”, “Photovoltaic identification”, etc.

This will allow AI to further penetrate into the fields, greatly improving the analysis efficiency of remote sensing applications such as disaster prevention, natural resource management, and agricultural yield estimation.

At present, this AI model has been applied in the industry. For example, the Shandong Provincial Land Surveying and Mapping Institute and Alibaba Damo Institute cooperated to use a large remote sensing AI model to monitor the growth of winter wheat. The recognition accuracy reached more than 90%, effectively improving the remote sensing interpretation of winter wheat. The efficiency can help agricultural managers better predict grain output and improve agricultural production efficiency.

The National Institute of Natural Disaster Prevention and Control also uses this model to identify landslides and collapsed buildings. In tests of remote sensing images of historical natural disaster areas, it only takes ten minutes to extract disaster information, which is faster than manual identification. The efficiency is dozens of times higher, providing efficient and accurate remote sensing analysis support for scientific disaster relief.

The above is the detailed content of DAMO Academy releases the industry's first large-scale remote sensing AI model, claiming to be able to identify nearly 100 types of land object classifications. For more information, please follow other related articles on the PHP Chinese website!