Hello Alexey,I am sorry that this is not something I want...I have 500 images for reconstructing a mesh and I have extra 10k images which I want to register. I just want to obtain the camera poses of the extra 10k images.It seems that "Incremental Image Alignment" tries to align inside those 10k images, which is extremely slow and causes out of memory failures.Is it possible to just detect the features from the extra 10k images and align them with the 500 images?Thanks!JianWang