OpenCodePapers

image-reconstruction-on-imagenet

Image Reconstruction
Dataset Link
Results over time
Click legend items to toggle metrics. Hover points for model names.
Leaderboard
Click a sortable column header to sort.
PaperCodeFIDLPIPSPSNRSSIMModelNameReleaseDate
MGVQ: Could VQ-VAE Beat VAE? A Generalizable Tokenizer with Multi-group Quantization✓ Link0.490.08624.700.787MGVQ (16x16x8)2025-07-14
MGVQ: Could VQ-VAE Beat VAE? A Generalizable Tokenizer with Multi-group Quantization✓ Link0.640.11023.710.755MGVQ (16x16x4)2025-07-14
GigaTok: Scaling Visual Tokenizers to 3 Billion Parameters for Autoregressive Image Generation✓ Link0.790.194721.650.699GigaTok-XL-XXL2025-04-11
Preventing Local Pitfalls in Vector Quantization via Optimal Transport✓ Link0.910.06627.570.729OptVQ (16x16x8)2024-12-19
Preventing Local Pitfalls in Vector Quantization via Optimal Transport✓ Link1.000.07626.590.717OptVQ (16x16x4)2024-12-19
Taming Scalable Visual Tokenizer for Autoregressive Image Generation✓ Link1.000.2030IBQ (16x16)2024-12-03
MoVQ: Modulating Quantized Vectors for High-Fidelity Image Generation✓ Link1.120.11322.420.673Mo-VQGAN (16x16x4)2022-09-19
Open-MAGVIT2: An Open-Source Project Toward Democratizing Auto-regressive Visual Generation✓ Link1.1721.90Open-Magvit2 (16x16)2024-09-06
Vector-quantized Image Modeling with Improved VQGAN✓ Link1.28ViT-VQGAN (16x16)2021-10-09
MaskBit: Embedding-free Image Generation via Bit Tokens✓ Link1.66MaskBit (16x16)2024-09-24
An Image is Worth 32 Tokens for Reconstruction and Generation✓ Link1.71TiTok-S-1282024-06-11
Autoregressive Image Generation using Residual Quantization✓ Link1.83RQ-VAE (8x8x16)2022-03-03
MaskGIT: Masked Generative Image Transformer✓ Link2.28MaskGIT-VQGAN (16x16)2022-02-08
Scaling the Codebook Size of VQGAN to 100,000 with a Utilization Rate of 99%✓ Link2.620.12023.800.589VQGAN-LC (16x16)2024-06-17
Taming Transformers for High-Resolution Image Synthesis✓ Link3.640.17719.930.542Taming-VQGAN (16x16)2020-12-17
Check out how to contribute new results.