Open Michaelwhite34 opened 1 year ago
Yes, this is straightforward. You can adapt the included data readers to your datasets: https://github.com/NVlabs/nvdiffrecmc/tree/main/dataset
We require camera poses and foreground segmentation masks (stored in the alpha channel of the images), and I would recommend 50+ images with nicely distributed views covering the object.
Yes, this is straightforward. You can adapt the included data readers to your datasets: https://github.com/NVlabs/nvdiffrecmc/tree/main/dataset
We require camera poses and foreground segmentation masks (stored in the alpha channel of the images), and I would recommend 50+ images with nicely distributed views covering the object.