There are many paths to take in processing data sets, depending on your end goal and available data

Full Metadata

Building a Voxel World and Visualizing given images and KRT Cameras. This tutorial will use the Downtown Providence data (images and cameras).

1. Image Upload
2. Viewing Images and Visualizing Cameras (Optional)
3. Building a Voxel World
4. Improving voxel world model (Optional)
5. Ordering a Point Cloud
6. Viewing a Point Cloud (Optional)
7. Error Analysis (Optional)
8. Download data (Optional)

Building a world with partial metadata (GPS Only)

Most metadata formats supported do not have accurate enough orientation information to successfully build a 3D model without adjusting the cameras. In order to accomplish this, an SFM algorithm is used to accurately register the cameras. Available GPS data is used to georegister the model, but the orientations are initialized to a nadir view.

This tutorial can also use the Downtown Providence data (images and cameras), and just reregister the cameras by running SFM.

1. Image Upload
2. Ordering SFM
3. Viewing Images and Visualizing Cameras (Optional)
4. Building a Voxel World
5. Improving voxel world model (Optional)
6. Ordering a Point Cloud
7. Viewing a Point Cloud (Optional)
8. Error Analysis (Optional)
9. Download data (Optional)

No Metadata

When there is no metadata or the metadata is no good, there are other techniques to georegister a model using control points (such as survey points) and tie points to triangulate the model and reposition the camera models. This tutorial will walk you through this. The Downtown Providence data (images) can be used for this, with no cameras needed.

1. No Metadata Example