mlsanityData to 7 Oct 2026

Explicitly Minimizing the Blur Error of Variational Autoencoders

arXiv 2304.05939 ↗First posted 12 Apr 2023cs.CV cs.LG eess.IVcited by 29

Not on the map. Why?

Abstract (arXiv)

Variational autoencoders (VAEs) are powerful generative modelling methods, however they suffer from blurry generated samples and reconstructions compared to the images they have been trained on. Significant research effort has been spent to increase the generative capabilities by creating more flexible models but often flexibility comes at the cost of higher complexity and computational cost. Several works have focused on altering the reconstruction term of the evidence lower bound (ELBO), however, often at the expense of losing the mathematical link to maximizing the likelihood of the samples under the modeled distribution. Here we propose a new formulation of the reconstruction term for the VAE that specifically penalizes the generation of blurry images while at the same time still maximizing the ELBO under the modeled distribution. We show the potential of the proposed loss on three different data sets, where it outperforms several recently proposed reconstruction losses for VAEs.

Key facts. Cited by 29 papers in the map (data through 7 Oct 2026). Builds on 21 of its 31 references that are in the map. First posted on arXiv 12 Apr 2023.

Builds on

21 of 31 references.

References · 21

  1. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  2. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  3. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  4. 2013Auto-Encoding Variational Bayescited by 3,980
  5. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  6. 2015Variational Inference with Normalizing Flowscited by 2,161
  7. 2019An Introduction to Variational Autoencoderscited by 997
  8. 2016Improving Variational Inference with Inverse Autoregressive Flowcited by 636
  9. 2020NVAE: A Deep Hierarchical Variational Autoencodercited by 635
  10. 2015Autoencoding beyond pixels using a learned similarity metriccited by 407
  11. 2017A General and Adaptive Robust Loss Functioncited by 236
  12. 2019Stochastic Latent Actor-Critic: Deep Reinforcement Learning with a Latent Variable Modelcited by 228
9 more
  1. 2020Focal Frequency Loss for Image Reconstruction and Synthesiscited by 204
  2. 2020Flows for simultaneous manifold learning and density estimationcited by 117
  3. 2019Improved Conditional VRNNs for Video Predictioncited by 104
  4. 2018Sylvester Normalizing Flows for Variational Inferencecited by 71
  5. 2021Rectangular Flows for Manifold Learningcited by 49
  6. 2018Blind Image Deconvolution using Deep Generative Priorscited by 40
  7. 2020A Loss Function for Generative Neural Networks Based on Watson's Perceptual Modelcited by 39
  8. 2017MR image reconstruction using deep density priorscited by 37
  9. 2020Modelling the Distribution of 3D Brain MRI using a 2D Slice VAEcited by 20

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 1; 2024: 10; 2025: 13; 2026: 5.

29 papers.

Sort:

By area:

  1. 2024PID: Physics-Informed Diffusion Model for Infrared Image Generationcited by 17
  2. 2025Flow Matching Meets Biology and Life Science: A Surveycited by 14
  3. 2023Tree Variational Autoencoderscited by 10
  4. 2025FELLE: Autoregressive Speech Synthesis with Token-Wise Coarse-to-Fine Flow Matchingcited by 8
  5. 2024BiPO: Bidirectional Partial Occlusion Network for Text-to-Motion Synthesiscited by 6
  6. 2024Diffusion Prior Interpolation for Flexibility Real-World Face Super-Resolutioncited by 5
  7. 2025Synergy Between Sufficient Changes and Sparse Mixing Procedure for Disentangled Representation Learningcited by 4
  8. 2024Blind Image Deconvolution by Generative-based Kernel Prior and Initializer via Latent Encodingcited by 3
  9. 2024GUIDE-VAE: Advancing Data Generation with User Information and Pattern Dictionariescited by 3
  10. 2024Text-to-Events: Synthetic Event Camera Streams from Conditional Text Inputcited by 2
  11. 2024TreeDiffusion: Hierarchical Generative Clustering for Conditional Diffusioncited by 2
  12. 2025Variational Rank Reduction Autoencoderscited by 2
17 more
  1. 2025Guided Unconditional and Conditional Generative Models for Super-Resolution and Inference of Quasi-Geostrophic Turbulencecited by 2
  2. 2024Short-Time Fourier Transform for deblurring Variational Autoencoderscited by 1
  3. 2024Motion Free B-frame Coding for Neural Video Compressioncited by 1
  4. 2026Diffusion-Based Authentication of Copy Detection Patterns: A Multimodal Framework with Printer Signature Conditioningcited by 1
  5. 2026Cross Modality Image Translation In Medical Imaging Using Generative Frameworkscited by 1
  6. 2026Content-Style Identification via Differential Independencecited by 1
  7. 2024Structured Generations: Using Hierarchical Clusters to guide Diffusion Modelscited by 0
  8. 2025A new pathway to generative artificial intelligence by minimizing the maximum entropycited by 0
  9. 20253D Nephrographic Image Synthesis in CT Urography with the Diffusion Model and Swin Transformercited by 0
  10. 2025Likelihood-Free Variational Autoencoderscited by 0
  11. 2025Learning Extrapolative Sequence Transformations from Markov Chainscited by 0
  12. 2025Generative Latent Kernel Modeling for Blind Motion Deblurringcited by 0
  13. 2025Structure-Preserving Unpaired Image Translation to Photometrically Calibrate JunoCam with Hubble Datacited by 0
  14. 2025mViSE: A Visual Search Engine for Analyzing Multiplex IHC Brain Tissue Imagescited by 0
  15. 2025Unified Control for Inference-Time Guidance of Denoising Diffusion Modelscited by 0
  16. 2026Deep Neural Networks as Iterated Function Systems and a Generalization Boundcited by 0
  17. 2026Rare-Aware Autoencoding: Reconstructing Spatially Imbalanced Datacited by 0