How did you guys come up with such an ingenious idea? #2
Description
Delved deep into your paper. The GaussianIP algorithm you proposed represents a remarkable leap forward, effectively overcoming issues like extended training durations and insufficient facial and garment details plaguing current 3D human generation methods.
Really curious about the thought trajectory that led to its creation. What initially spurred you to develop the Adaptive Human Distillation Sampling (AHDS) method in the first stage? How did you determine that leveraging human - centric knowledge would streamline the generation process? Additionally, could you share the reasoning behind designing the View - Consistent Refinement (VCR) strategy for enhancing face and clothing details in the second stage? Were there any particular challenges during implementation that steered your design decisions?