Digital archiving cultural properties using 3D models can provide a more intuitive understanding of the 3D geometry. However, high-quality 3D model reconstruction using laser scanning is expensive and time-consuming. An image-based method like Structure from Motion (SfM) is a more cost-effective approach. Nevertheless, SfM requires many multi-view images to produce reliable results. Although the crowdsourcing approach can be applied to gather a large amount of image data from visitors, the quality of the data cannot be guaranteed. In this paper, we propose an image-gathering method upon crowdsourcing. Our method contains a VR-assisted interface that the experts use to generate snapshot instructions for visitors to improve data quality.