Towards image compression with perfect realism at ultra-low bitrates

M Careil, MJ Muckley, J Verbeek… - The Twelfth International …, 2023 - openreview.net
Image codecs are typically optimized to trade-off bitrate vs. distortion metrics. At low bitrates,
this leads to compression artefacts which are easily perceptible, even when training with …

Learning to generate semantic layouts for higher text-image correspondence in text-to-image synthesis

M Park, J Yun, S Choi, J Choo - Proceedings of the IEEE …, 2023 - openaccess.thecvf.com
Existing text-to-image generation approaches have set high standards for photorealism and
text-image correspondence, largely benefiting from web-scale text-image datasets, which …

Modulating pretrained diffusion models for multimodal image synthesis

C Ham, J Hays, J Lu, KK Singh, Z Zhang… - ACM SIGGRAPH 2023 …, 2023 - dl.acm.org
We present multimodal conditioning modules (MCM) for enabling conditional image
synthesis using pretrained diffusion models. Previous multimodal synthesis works rely on …

Improving statistical fidelity for neural image compression with implicit local likelihood models

MJ Muckley, A El-Nouby, K Ullrich… - International …, 2023 - proceedings.mlr.press
Lossy image compression aims to represent images in as few bits as possible while
maintaining fidelity to the original. Theoretical results indicate that optimizing distortion …

Diagen: diverse image augmentation with generative models

T Lingenberg, M Reuter, G Sudhakaran… - arxiv preprint arxiv …, 2024 - arxiv.org
Simple data augmentation techniques, such as rotations and flips, are widely used to
enhance the generalization power of computer vision models. However, these techniques …

Utilizing greedy nature for multimodal conditional image synthesis in transformers

S Su, J Zhu, L Gao, J Song - IEEE Transactions on Multimedia, 2023 - ieeexplore.ieee.org
Multimodal Conditional Image Synthesis (MCIS) aims to generate images according to
different modalities input and their combination, which allows users to describe their …

Location-aware adaptive normalization: a deep learning approach for wildfire danger forecasting

MHS Eddin, R Roscher, J Gall - IEEE Transactions on …, 2023 - ieeexplore.ieee.org
Climate change is expected to intensify and increase extreme events in the weather cycle.
Since this has a significant impact on various sectors of our life, recent works are concerned …

medigan: a Python library of pretrained generative models for medical image synthesis

R Osuala, G Skorupko, N Lazrak… - Journal of Medical …, 2023 - spiedigitallibrary.org
Purpose Deep learning has shown great promise as the backbone of clinical decision
support systems. Synthetic data generated by generative models can enhance the …

UNet-like network fused swin transformer and CNN for semantic image synthesis

A Ke, J Luo, B Cai - Scientific Reports, 2024 - nature.com
Semantic image synthesis approaches has been dominated by the modelling of
Convolutional Neural Networks (CNN). Due to the limitations of local perception, their …

Adversarial supervision makes layout-to-image diffusion models thrive

Y Li, M Keuper, D Zhang, A Khoreva - The Twelfth International …, 2024 - openreview.net
Despite the recent advances in large-scale diffusion models, little progress has been made
on the layout-to-image (L2I) synthesis task. Current L2I models either suffer from poor …