Test Dataset of AI Altered Images for Photogrammetric Reconstruction

As a part of a test of the capability of using AI altered images in photogrammetry, I generated these image sets using the Prisma AI app. All of these datasets will result in a model in Agisoft Metashape. Prisma AI acts as a filter, changing an input image to the style of another artwork. 

Original images were collected using a Samsung S22 camera on the 1x zoom rear lens. The subject is a rock decoration in the Utrecht University Botanical Garden.

Results can be seen on Sketchfab: https://sketchfab.com/gspeed0689/collections/prisma-photogrammetry-2e334a36c6e942798370d594de5e6f0c 

Included datasets:

true-color.zip - The original images from the Samsung S22

aviator-lowres - Processed images at normal (low resolution) quality with the filter labelled Aviator

bubblegum-lowres - Processed images at normal (low resolution) quality with the filter labelled Bubblegum

gothic-lowres - Processed images at normal (low resolution) quality with the filter labelled Gothic

gothic-hd.zip - Processed images at HD (paid subscription required) quality with the filter labelled Gothic

true-color.fbx - 3D model created with Agisoft Metashape using the parameters listed below and the images from true-color.zip

aviator-lowres.fbx - 3D model created with Agisoft Metashape using the parameters listed below and the images from aviator-lowres.zip

bubblegum-lowres.fbx - 3D model created with Agisoft Metashape using the parameters listed below and the images from bubblegum-lowres.zip

gothic-lowres.fbx - 3D model created with Agisoft Metashape using the parameters listed below and the images from gothic-lowres.zip

gothic-hd.fbx - 3D model created with Agisoft Metashape using the parameters listed below and the images from gothic-hd.zip

Processing parameters in Agisoft Metashape v 1.7:

Alignment defaults

Dense cloud creation at "High" quality

Mesh defaults