Hello Everyone,
Had a conceptual question regarding the evaluation of diffusion models.
Are there quantitative ways to measure the performance of the sampled images from the trained network ? So, when are model is generating new images (with new embeddings), how can we quantitatively describe the performance?