A Complex Dataset for Photogrammetry & View Simulation Processing
Mike Heath Mike Heath
134 subscribers
135 views
0

 Published On Oct 24, 2023

I review a recent dataset of imagery I captured with an action camera for use in 3D modeling with Photogrammetry and view simulation with radiance fields. Neural Radiance Fields (NeRFs) & Gaussian Splatting are two examples of the latter (but I don't show results from those methods in this video). The dataset is interesting because its so geometrically complex and because it was captured from many camera perspectives. It was captured as video with an wide angle action camera (Insta360 One-R 1 inch) mounted on a pole that was carried around the campsite (a drone was not used). A virtual fly-through of a 3D vertex colored mesh model illustrates the camera positions at the extracted frame times (one frame per second). It was created with Agisoft Metashape using traditional photogrammetry. Multiple issues such as sun flares through the lens, people moving around, people moving some objects around the camp and thin surfaces (tent canopies) pose challenges to photogrammetry processing. This could be a great dataset for use in processing with different algorithms and approaches. Please contact me (see the about section of my channel) if you would like to use this imagery dataset.

show more

Share/Embed