Scaling Image Tokenizers with Grouped Spherical Quantization
This work addresses scalability and efficiency issues in vision tokenizers for researchers and practitioners in computer vision, though it appears incremental as it builds on existing GAN-based methods.
The paper tackled the problem of scaling image tokenizers by introducing Grouped Spherical Quantization (GSQ), which achieved a 16x down-sampling with a reconstruction FID of 0.50, outperforming state-of-the-art methods in reconstruction quality with fewer training iterations.
Vision tokenizers have gained a lot of attraction due to their scalability and compactness; previous works depend on old-school GAN-based hyperparameters, biased comparisons, and a lack of comprehensive analysis of the scaling behaviours. To tackle those issues, we introduce Grouped Spherical Quantization (GSQ), featuring spherical codebook initialization and lookup regularization to constrain codebook latent to a spherical surface. Our empirical analysis of image tokenizer training strategies demonstrates that GSQ-GAN achieves superior reconstruction quality over state-of-the-art methods with fewer training iterations, providing a solid foundation for scaling studies. Building on this, we systematically examine the scaling behaviours of GSQ, specifically in latent dimensionality, codebook size, and compression ratios, and their impact on model performance. Our findings reveal distinct behaviours at high and low spatial compression levels, underscoring challenges in representing high-dimensional latent spaces. We show that GSQ can restructure high-dimensional latent into compact, low-dimensional spaces, thus enabling efficient scaling with improved quality. As a result, GSQ-GAN achieves a 16x down-sampling with a reconstruction FID (rFID) of 0.50.