1. Chinese Ancient Architecture Image Database

We have captured typical ancient architectures in the four Buddhist mountains (Mount Wutai, Mount Emei, Mount Jiuhua, Mount Puto) and two Taoist mountains (Mount Wudang, Mount Qingcheng). The camera is Canno EOS 5D. More information you can visit

1.1 Benchmark Database

We use the Riegl-LMS-Z420i laser scanner to scan the 3 buildings in the Tsinghua University. The scanner's accuracy is 10mm@50m, and the angular step width is 0.0057degree. The 3 buildings are the old gate, the Tsinghua school, and the life science building. You can download the ground truth data from laser scanner, the image data and the camera projection matrix of each buildings. And you can also obtain the evaluation software which is used to evaluate the difference between your result and the ground truth in Software section. More information you can visit

1.2 Benchmark Data Evaluation Software

You can download the Benchmark Data Evaluation Software written by Wei Gao to evaluate the difference between your result and the ground truth.

 The software is used to compare your results of 3d reconstruction with the groundtruth data. You can obtain the accuracy and completeness similar with the Seitz's scheme*. We now release the 64bit Windows Editon.
*: Steve Seitz, Brian Curless, etc., "A Comparison and Evaluation of Multi-View Stereo Reconstruction Algorithms", CVPR 2006.

The latest version of Evaluation is 0.2. Evaluation is distributed in Windows 64bit binary format.

  • Evaluation Version 0.2 (released July 12, 2012) (Change log)  [Download]
    [Note: Sorry for cannot download the software, as we have found some bug about reading wrml file, we will fix it soon later. Please keep on watching new version. Thanks! ]
  • Evaluation Version 0.1 (released February 4, 2011)

2. A large scale dataset for indoor visual localization with high precise ground truth

This paper presents a challenging new dataset for indoor localization research. We have recorded the whole internal structure of Fengtai Wanda Plaza which is over 15800 square meters area with a Navvis M6 device. The dataset contains over one thousand RGBD panoramas and three thousand query images collected by three different smartphones. In addition to the data, aligned 3D point cloud is produced after the elimination of moving objects based on the building floorplan. Furthermore, a method is provided to generate corresponding high resolution depth images for each panorama. By fixing the smartphones on the device using a specially designed bracket, 6 degree of freedom camera poses can be calculated precisely. We believe it can give a new benchmark for indoor visual localization. More information you can visit [Here].