扩散模型生成的合成数据提升 ImageNet 分类准确性（英文版）

英文标题：Synthetic Data from Diffusion Models Improves ImageNet Classification中文摘要：本研究使用大规模的文本到图像扩散模型对分类条件模型进行微调，进而在 ImageNet 分类准确性得分上实现了显著的提升，证明了利用自然图像模型进行生成数据增强的可行性。英文摘要：Deep generative models are becoming increasingly powerful, now generatingdiverse high fidelity photo-realistic samples given text prompts. Have theyreached the point where models of natural images can be used for generativedata augmentation, helping to improve challenging discriminative tasks? We showthat large-scale text-to image