Structure-from-Motion Alignment Dataset

Dataset Overview

Name: Structure‑from‑Motion Alignment Dataset

Purpose: Evaluate algorithms' ability to calibrate two Structure‑from‑Motion reconstructions under unknown relative pose and scale.

Construction Method:

• Recorded multiple image sequences in a controlled indoor environment.
• Used an ART‑2 tracking system to precisely track markers attached to the camera.
• Processed a long image sequence with an offline incremental SfM pipeline to generate a scene point‑cloud reconstruction and calibrate the transformation between camera and tracker coordinate systems.
• Recorded twelve image sequences in the tracked environment and processed them with a real‑time keyframe‑based SLAM system.

Data Format:

• All data stored as binary files containing float or double values.
• File extensions indicate format, e.g., double3 for a double‑precision 3‑vector.
• Top‑level includes descriptors and 3D points with their descriptors.
• Each sub‑folder represents an image sequence; each keyframe contains descriptors, features (2D points), ground‑truth camera pose, and SLAM‑estimated camera pose.

License:

• The dataset is released under the Creative Commons Attribution 4.0 International License.

Citation Information

If you use this dataset, please cite the following papers:

1. Ventura, J., Arth, C., Reitmayr, G., & Schmalstieg, D. (2014). Global Localization from Monocular SLAM on a Mobile Phone. IEEE Transactions on Visualization and Computer Graphics, 2014.
2. Ventura, J., Arth, C., Reitmayr, G., & Schmalstieg, D. (2014). A minimal solution to the generalized pose‑and‑scale problem. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.