Deep compression autoencoder boosts image generation speed and quality
DC-SAE: Deep Compression Semantic Autoencoder for Faster Diffusion Convergence
Computer Vision and Pattern Recognition
Summary
Generating high-quality images with AI usually takes a lot of training time, especially when compressing images into small representations. The authors present DC-SAE, a new technique that compresses images more while keeping important details, making AI models learn faster and create better images. Their method combines a high-level semantic encoder with a detail-preserving encoder to balance compression and image quality. They show it works well on large image datasets and helps create high-resolution images from text prompts more efficiently.
What this means in practice
- •For image generation engineers: Train diffusion-based image generation systems faster while preserving image quality through improved compression and encoding techniques.
- •For text-to-image system developers: Build efficient text-to-image generation tools that handle high-resolution outputs with faster convergence and better image fidelity.
Authors
Xu Huang, Ye Huang, Zijun Liao, Yuwei Niu, Xiaojie Li, Menghan Zhou, De Wen Soh, Xiaotong Li, Daquan Zhou
Abstract
High-compression tokenizers are essential for scaling latent image generative models. However, aggressive compression creates a fundamental tradeoff between reconstruction fidelity and generation efficiency: high compression image encoder always increases the learning difficulty of diffusion training, resulting in slow model convergence. Recent representation autoencoders speed up the diffusion training by improving the latent feature's expressive capability by replacing VAE encoders with pretrained semantic encoders, yet they are typically limited to moderate compression and lose pixel-level details necessary for faithful reconstruction. To achieve both high compression and fast diffusion training, we propose DC-SAE, a Decoupled Compact Semantic Autoencoder designed for high-compression image generation with accelerated diffusion model convergence. DC-SAE consists of two key components: (1) a macro-level architecture design that leverages semantic encoders to enable higher compression ratios, and (2) a pixel-level encoder that preserves low-level details, ensuring high-fidelity image reconstruction. We empirically demonstrate that DC-SAE performs strongly on image generation tasks, achieving both compact latent representations and efficient training dynamics. Specifically, on the ImageNet dataset with $512 \times 512$ resolution, DC-SAE achieves $32\times$ spatial compression, with 29.79 PSNR and 3.37 gFID, substantially outperforming the previous state-of-the-art high-compression tokenizer baselines DC-AE by 13.5% and 54.9% on PSNR and gFID, respectively, maintaining comparable throughput and faster diffusion model training convergence. Beyond class-conditional generation, a $1.6$B-parameter DiT using DC-SAE achieves 0.84 on GenEval and 86.007 on DPG-Bench for text-to-image generation at $1024\times1024$ resolution.