Research map / Image generation / Autoregressive and latent generation
MoVQ: Modulating Quantized Vectors for High-Fidelity Image Generation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace GroupNorm with spatially conditional normalization plus multichannel quantization (c=4) in VQGAN decoder | VQGAN · FFHQ reconstruction | rFID: 4.42 -> 2.26 | Helped |
| same, on ImageNet with 1024 codes | VQGAN · ImageNet reconstruction | rFID: 6.25 -> 1.12 | Helped |
| use MoVQ quantizer for second-stage generation | VQGAN · ImageNet class-conditional | FID: 15.78 -> 7.13 | Helped |
| use MoVQ quantizer for unconditional generation | VQGAN · FFHQ unconditional | FID: 11.4 -> 8.52 | Helped |
| add multichannel representation to VQGAN (ablation) | VQGAN · FFHQ reconstruction | reconstruction quality | Helped |
| add spatial normalization with Fourier-feature initial input (ablation) | MoVQ multichannel · FFHQ reconstruction | reconstruction quality | Helped |
| use positional embedding or learned constant as initial feature in spatial normalization (ablation) | MoVQ multichannel · FFHQ reconstruction | reconstruction quality | No clear effect |
| increase codebook from 1024 to 16384 (ablation) | MoVQ · FFHQ reconstruction | reconstruction quality | No clear effect |
| more latent channels (ablation) | MoVQ · generation | generation quality | Mixed |
| masked parallel sampling (8-12 steps) instead of autoregressive (1024 steps) | Mo-VQGAN · ImageNet class-conditional | FID: 7.13 -> 7.22 | No clear effect |
Closest papers by meaning
- Rethinking the Objectives of Vector-Quantized Tokenizers for Image Synthesis (CVPR 2024 · similarity 0.90)
- Efficient Generative Modeling with Residual Vector Quantization-Based Tokens (ICML 2025 · similarity 0.88)
- Finite Scalar Quantization: VQ-VAE Made Simple (ICLR 2024 · similarity 0.88)
- HyperVQ: Enabling Hyperprior Entropy Modeling for VQ-Based Generative Image Compression (NeurIPS 2026 · similarity 0.88)
- SoftVQ-VAE: Efficient 1-Dimensional Continuous Tokenizer (CVPR 2025 · similarity 0.87)
- Learning to Quantize for Training Vector-Quantized Networks (ICML 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.