Filters








140,190 Hits in 7.7 sec

Image Generation with Multimodal Priors using Denoising Diffusion Probabilistic Models [article]

Nithin Gopalakrishnan Nair, Wele Gedara Chaminda Bandara, Vishal M Patel
2022 arXiv   pre-print
To this end, we propose a solution based on a denoising diffusion probabilistic models to synthesise images under multi-model priors.  ...  Based on the fact that the distribution over each time step in the diffusion model is Gaussian, in this work we show that there exists a closed-form expression to the generate the image corresponds to  ...  Recently diffusion models thrived over other generative models for the task of image generation [5] .  ... 
arXiv:2206.05039v1 fatcat:whb7qq5etzbm5jvmskb43n3k2m

A Novel Unified Conditional Score-based Generative Framework for Multi-modal Medical Image Completion [article]

Xiangxi Meng, Yuning Gu, Yongsheng Pan, Nizhuan Wang, Peng Xue, Mengkang Lu, Xuming He, Yiqiang Zhan, Dinggang Shen
2022 arXiv   pre-print
Here, we propose the Unified Multi-Modal Conditional Score-based Generative Model (UMM-CSGM) to take advantage of Score-based Generative Model (SGM) in modeling and stochastically sampling a target probability  ...  Specifically, UMM-CSGM employs a novel multi-in multi-out Conditional Score Network (mm-CSN) to learn a comprehensive set of cross-modal conditional distributions via conditional diffusion and reverse  ...  Unified multi-modal conditional score-based generative model Naive diffusion and score-based reverse generation.  ... 
arXiv:2207.03430v1 fatcat:cxt2a5574beh3p3njtx3eyhyue

ProstateGAN: Mitigating Data Bias via Prostate Diffusion Imaging Synthesis with Generative Adversarial Networks [article]

Xiaodan Hu, Audrey G. Chung, Paul Fieguth, Farzad Khalvati, Masoom A. Haider, Alexander Wong
2018 arXiv   pre-print
Motivated by this, in this study we introduce ProstateGAN, a GAN-based model for synthesizing realistic prostate diffusion imaging data.  ...  More specifically, in order to generate new diffusion imaging data corresponding to a particular cancer grade (Gleason score), we propose a conditional deep convolutional GAN architecture that takes Gleason  ...  In this paper, we propose ProstateGAN, a GAN-based model for synthesizing prostate diffusion imaging data, which can be used to mitigate the data bias present in machine learning-driven prostate cancer  ... 
arXiv:1811.05817v2 fatcat:qd3maf6earcjbngdinlint6xue

Cascaded Diffusion Models for High Fidelity Image Generation [article]

Jonathan Ho, Chitwan Saharia, William Chan, David J. Fleet, Mohammad Norouzi, Tim Salimans
2021 arXiv   pre-print
A cascaded diffusion model comprises a pipeline of multiple diffusion models that generate images of increasing resolution, beginning with a standard diffusion model at the lowest resolution, followed  ...  We show that cascaded diffusion models are capable of generating high fidelity images on the class-conditional ImageNet generation benchmark, without any assistance from auxiliary image classifiers to  ...  Conclusion We have shown that cascaded diffusion models are capable of outperforming state-of-the-art generative models on the ImageNet class-conditional generation benchmark when paired with conditioning  ... 
arXiv:2106.15282v3 fatcat:f3mrzyjv5jcndjdg6icfcs6zrq

Improved Vector Quantized Diffusion Models [article]

Zhicong Tang, Shuyang Gu, Jianmin Bao, Dong Chen, Fang Wen
2022 arXiv   pre-print
Vector quantized diffusion (VQ-Diffusion) is a powerful generative model for text-to-image synthesis, but sometimes can still generate low-quality samples or weakly correlated images with text input.  ...  We achieve an 8.44 FID score on MSCOCO, surpassing VQ-Diffusion by 5.42 FID score.  ...  Thus, we find the VQ-Diffusion may easily generate images with poor correlation with input text, which is often calculated with CLIP score [32] .  ... 
arXiv:2205.16007v1 fatcat:peq6hg4vhfdbnderwkzl7qok7m

Discrete Contrastive Diffusion for Cross-Modal and Conditional Generation [article]

Ye Zhu, Yu Wu, Kyle Olszewski, Jian Ren, Sergey Tulyakov, Yan Yan
2022 arXiv   pre-print
We demonstrate the efficacy of our approach in evaluations with three diverse, multimodal conditional synthesis tasks: dance-to-music generation, text-to-image synthesis, and class-conditioned image synthesis  ...  Diffusion probabilistic models (DPMs) have become a popular approach to conditional generation, due to their promising results and support for cross-modal synthesis.  ...  Compared to the unconditional case, conditional generation is usually applied in more concrete and practical cross-modality scenarios, e.g. video-based music generation [17, 90, 23] and text-based image  ... 
arXiv:2206.07771v1 fatcat:hijvmzwuzndzrfscwa7agworzi

Diffusion models for Handwriting Generation [article]

Troy Luhman, Eric Luhman
2020 arXiv   pre-print
In this paper, we propose a diffusion probabilistic model for handwriting generation.  ...  Diffusion models are a class of generative models where samples start from Gaussian noise and are gradually denoised to produce output.  ...  Score-based generative models estimate the gradient of the logarithmic data density ∇ x log (p (x)) with a denoising objective.  ... 
arXiv:2011.06704v1 fatcat:2mgqwpkqbrab3g44uawh3ojqaa

Compositional Visual Generation with Composable Diffusion Models [article]

Nan Liu, Shuang Li, Yilun Du, Antonio Torralba, Joshua B. Tenenbaum
2022 arXiv   pre-print
An image is generated by composing a set of diffusion models, with each of them modeling a certain component of the image.  ...  Project page: https://energy-based-model.github.io/Compositional-Visual-Generation-with-Composable-Diffusion-Models/  ...  Shuang Li is supported by Raytheon BBN Technologies Corp. under the project Symbiant (reg. no. 030256-00001 90113), Mitsubishi Electric Research Laboratory (MERL) under the project Generative Models For  ... 
arXiv:2206.01714v4 fatcat:t43zywkofzcuhnhv4blfb5ikse

DiffuseMorph: Unsupervised Deformable Image Registration Along Continuous Trajectory Using Diffusion Models [article]

Boah Kim, Inhwa Han, Jong Chul Ye
2021 arXiv   pre-print
To address this, here we present a novel approach of diffusion model-based probabilistic image registration, called DiffuseMorph.  ...  Specifically, our model learns the score function of the deformation between moving and fixed images.  ...  To generate images with desired semantics, conditional denoising diffusion models have been also presented [9, 34] .  ... 
arXiv:2112.05149v1 fatcat:cdktrtbixbalhmsvukrtjrdzu4

Non-Uniform Diffusion Models [article]

Georgios Batzolis, Jan Stanczuk, Carola-Bibiane Schönlieb, Christian Etmann
2022 arXiv   pre-print
Moreover, we show that non-uniform diffusion leads to a novel estimator for the conditional score function which achieves on par performance with the state-of-the-art conditional denoising estimator.  ...  Diffusion models have emerged as one of the most promising frameworks for deep generative modeling. In this work, we explore the potential of non-uniform diffusion models.  ...  Conditional Generation In this section we conduct a systematic comparison of different score-based diffusion approaches to modelling conditional distributions of image data.  ... 
arXiv:2207.09786v1 fatcat:7nnx2r64pved7dabwivmvvc2jm

Pyramidal Denoising Diffusion Probabilistic Models [article]

Dohoon Ryu, Jong Chul Ye
2022 arXiv   pre-print
To address this problem, here we present a novel pyramidal diffusion model to generate high resolution images starting from much coarser resolution images using a single score function trained with a positional  ...  Diffusion models have demonstrated impressive image generation performance, and have been used in various computer vision tasks.  ...  Denoising diffusion probabilistic model (DDPM) [10, 37] can be considered as discrete form of score-based generative models.  ... 
arXiv:2208.01864v1 fatcat:cd2ikjwwxvd53ebxqd7jdenrgq

KNN-Diffusion: Image Generation via Large-Scale Retrieval [article]

Oron Ashual, Shelly Sheynin, Adam Polyak, Uriel Singer, Oran Gafni, Eliya Nachmani, Yaniv Taigman
2022 arXiv   pre-print
Our diffusion-based model trains on images only, by leveraging a joint Text-Image multi-modal metric.  ...  Compared to baseline methods, our generations achieve state of the art results both in human evaluations as well as with perceptual scores when tested on a public multimodal dataset of natural images,  ...  We train a diffusion-based generative model G, with parameters θ, which instead of conditioning it on f txt (t), fetches the K nearest image embed- dings {f img (I i k )} K k=1 of f img (I) in a joint  ... 
arXiv:2204.02849v1 fatcat:usyym6vikjhvll5kroadghg7ai

Diffusion Autoencoders: Toward a Meaningful and Decodable Representation [article]

Konpat Preechakul, Nattanat Chatthee, Suttisak Wizadwongsa, Supasorn Suwajanakorn
2022 arXiv   pre-print
Diffusion probabilistic models (DPMs) have achieved remarkable quality in image generation that rivals GANs'.  ...  This capability enables challenging applications that currently foil GAN-based methods, such as attribute manipulation on real images.  ...  Related work Denoising diffusion-based generative models [22, 46] are closely related to denoising score-based generative models [48] .  ... 
arXiv:2111.15640v3 fatcat:sahzuednxbb4dpjxswv6yilx2i

Image Super-Resolution via Iterative Refinement [article]

Chitwan Saharia, Jonathan Ho, William Chan, Tim Salimans, David J. Fleet, Mohammad Norouzi
2021 arXiv   pre-print
SR3 adapts denoising diffusion probabilistic models to conditional image generation and performs super-resolution through a stochastic denoising process.  ...  We further show the effectiveness of SR3 in cascaded image generation, where generative models are chained with super-resolution models, yielding a competitive FID score of 11.3 on ImageNet.  ...  We also thank authors of [28] for generously providing us with baseline superresolution samples for human evaluation.  ... 
arXiv:2104.07636v2 fatcat:ae3bac4cyjgg3ayr2gdku2tq3e

Subspace Diffusion Generative Models [article]

Bowen Jing, Gabriele Corso, Renato Berlinghieri, Tommi Jaakkola
2022 arXiv   pre-print
Score-based models generate samples by mapping noise to data (and vice versa) via a high-dimensional diffusion process.  ...  Our framework is fully compatible with continuous-time diffusion and retains its flexible capabilities, including exact log-likelihoods and controllable generation.  ...  Hundreds or even thousands of evaluations of the high-dimensional score model are required to generate an image, making inference with score-based models much slower than with GANs or VAEs [6, 20] .  ... 
arXiv:2205.01490v1 fatcat:tfno2djfd5gp5feoprtg3c6cmy
« Previous Showing results 1 — 15 out of 140,190 results