EarthGPT: A Universal Multi-modal Large Language Model for Multi-sensor Image Comprehension in Remote Sensing Domain

Multi-modal large language models (MLLMs) have demonstrated remarkable success in vision and visual-language tasks within the natural image domain. Owing to the significant diversities between the natural image and RS image hinder the development of MLLMs in the remote sensing (RS) domain. Currently, the unified and powerful MLLM capable of various RS visual tasks is still under-explored. To fill the gap, a pioneer MLLM called EarthGPT is proposed for universal RS image comprehension, which integrates various multi-sensor RS interpretation tasks uniformly. More importantly, a large-scale multi-sensor multi-modal RS instruction-following dataset named MMRS is carefully constructed, which comprises 1005.842k image-text pairs based on 34 existing diverse RS datasets and includes multi-sensor images such as optical, synthetic aperture radar (SAR), and infrared. The MMRS addresses the issue of MLLMs lacking RS expert knowledge and stimulates the development of MMLMs in the RS domain. Extensive experiments demonstrate the EarthGPT's superior performance in various RS visual interpretation tasks compared with the other specialist models and MLLMs, which proves the effectiveness of the proposed EarthGPT and provides a versatile paradigm for open-set reasoning tasks.