ML Research Wiki / Benchmarks / Image Reconstruction / Ultra-High Resolution Image Reconstruction Benchmark

Ultra-High Resolution Image Reconstruction Benchmark

Image Reconstruction Benchmark

Performance Over Time

📊 Showing 6 results | 📏 Metric: rFID

Top Performing Models

Rank Model Paper rFID Date Code
1 MGVQ (16x16x4) MGVQ: Could VQ-VAE Beat VAE? A Generalizable Tokenizer with Multi-group Quantization 28.27 2025-07-14 📦 MKJia/MGVQ
2 SD-VAE (16x16) High-Resolution Image Synthesis with Latent Diffusion Models 26.86 2021-12-20 📦 compvis/stable-diffusion 📦 labmlai/annotated_deep_learning_paper_implementations 📦 stability-ai/stablediffusion
3 Open-Magvit2 (16x16) Open-MAGVIT2: An Open-Source Project Toward Democratizing Auto-regressive Visual Generation 23.91 2024-09-06 📦 tencentarc/open-magvit2 📦 tencentarc/seed-voken
4 LlamaGen (16x16) Autoregressive Model Beats Diffusion: Llama for Scalable Image Generation 23.90 2024-06-10 📦 foundationvision/llamagen 📦 0606zt/panollama
5 VQGAN (16x16) Taming Transformers for High-Resolution Image Synthesis 22.91 2020-12-17 📦 CompVis/taming-transformers 📦 alibaba/EasyNLP 📦 dome272/VQGAN
6 VAR (16x16) Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction 21.79 2024-04-03 📦 FoundationVision/VAR 📦 FoundationVision/Infinity 📦 EkaterinaXie/LiteVAR

All Papers (6)

MKJia/MGVQ
compvis/stable-diffusion labmlai/annotated_deep_learning_paper_implementations
tencentarc/open-magvit2 tencentarc/seed-voken
foundationvision/llamagen 0606zt/panollama
CompVis/taming-transformers alibaba/EasyNLP
FoundationVision/VAR FoundationVision/Infinity EkaterinaXie/LiteVAR