Reliable live load value is the prerequisite for reliability design of civil engineering buildings. Traditionally, building live loads are investigated by means of indoor objects sampling and weighing. This method has many problems such as low efficiency, high labor cost, long duration, limited samples, poor timelines in reflecting indoor items change and difficulty in weighing large items. Inspired by the big data concept, this paper proposes a new research method for investigating indoor sustained live loads. Through the multi-source heterogeneous data such as photos, videos, identification codes, and voices, combined with internet resources, the weight of objects in the building is obtained by means of object detection, image retrieval, voice or text recognition. After the detailed introduction of the implementation method, further case studies show that the use of big data technology can achieve efficient and convenient building live load survey, and build a new load research paradigm