What's new in DiverseShot V24?

Our V24 version includes significant improvements in user experience and stability. Highlights include a faster camera sparse reconstruction algorithm and support for training PNG image sets with transparent backgrounds.

V.2.4 UPDATES

Added three sparse reconstruction methods. Users can select the desired type in the UI ‘SfmSolver’, with Glomapper set as the default for the fastest performance. Sparse point clouds are displayed during the Sfm stage.

Introduced a Mask option to focus on training objects within the effective areas of images. If a mask folder is available, the corresponding mask files will be loaded. For PNG training images, the alpha channel will be used as the mask.

‍

‍

Added an option in GaussianComponent to exclude it from rendering. This can be set to false during training to speed up the process.

Other updates

1. Added a Resource Panel to view project resource files.
2. Added a ViewCube Axis.
3. Introduced Splat model merging and export functionality.
4. Fixed issues with Chinese path names.
5. Fixed the display issue of camera perspectives during training.

‍