feature volumes

This research introduces a 3D volumetric encoder for efficient and flexible text-to-3D generation. A lightweight network is developed to acquire feature volumes from multi-view images, which are then trained on…