TokenFlow: Unified Image Tokenizer for Multimodal Understanding and Generation
兼顾 Low-level 的 Pixel 信息和 High-level 的 Semantic 信息
CVPR
SoftVQ-VAE: Efficient 1-Dimensional Continuous Tokenizer
TiTok 框架 + Softmax 版 Vector Quantization 以期更高的压缩比
Autoregressive Image Generation using Residual Quantization
RQ-VAE, 残差向量量化
Taming Transformers for High-Resolution Image Synthesis
VQGAN, 自回归式的图片生成