New Features in Meshroom 2023.2 - alicevision/Meshroom GitHub Wiki

1. New Photometric Stereo nodes

There is a new Photometric Stereo pipeline where the depth map is obtained with several photographs taken from the same viewpoint, but under different lighting conditions - unlike the geometric approach, which uses a geometric model to derive depth information from the parallax available between acquisitions made from different viewpoints. For the current version, the inputs are the captured images with the direction and the intensity of the lights, and the outputs are the normal maps and the albedo. The light direction and intensity are provided with light balls - that need to be integrated in the shooting set. These light balls are then automatically detected in Meshroom with Machine Learning technique.

Input images need to be located in a folder that starts with ‘ps_’ to be grouped together using their pose ID. The photometric stereo cannot run otherwise. A sample dataset can be found here https://github.com/alicevision/dataset_photometricStereo

grafik

It is possible to use the <- and -> arrows in the node settings to position the circle on the sphere.

2. New Camera Tracking Pipeline

We introduced a new camera tracking pipeline with distortion calibration (calculated with checkerboard detection), keyframes selection to separate tracking into 2 reconstruction phases (increase robustness) and scene geometry estimation with exports and preview. This new pipeline can take advantage of a pre-existing photogrammetry of the shooting set.

The pipeline now has three CameraInit nodes, labelled as follows:

"InitPhotogrammetry", which is meant to be set with the images for the photogrammetry.
"InitLensGrid", which should be provided with the checkerboard (if it exists) for the distortion calibration. If no calibration grid is available, then the group of nodes (or the connection to ApplyCalibration) can be removed.
"InitShot", which is meant to be set with the images for the tracking.

3. New UI feature