While image generation with diffusion models has achieved a great success, generating images of higher resolution than the training size remains a challenging task due to the high computational cost. Current methods typically perform the entire sampling process at full resolution and process all frequency components simultaneously, contradicting with the inherent coarse-to-fine nature of latent diffusion models and wasting computations on processing premature high-frequency details at early diffusion stages. To address this issue, we introduce an efficient $\textbf{Fre}$quency-aware $\textbf{Ca}$scaded $\textbf{S}$ampling framework, $\textbf{FreCaS}$ in short, for higher-resolution image generation. FreCaS decomposes the sampling process into cascaded stages with gradually increased resolutions, progressively expanding frequency bands and refining the corresponding details. We propose an innovative frequency-aware classifier-free guidance (FA-CFG) strategy to assign different guidance strengths for different frequency components, directing the diffusion model to add new details in the expanded frequency domain of each stage. Additionally, we fuse the cross-attention maps of previous and current stages to avoid synthesizing unfaithful layouts. Experiments demonstrate that FreCaS significantly outperforms state-of-the-art methods in image quality and generation speed. In particular, FreCaS is about 2.86$\times$ and 6.07$\times$ faster than ScaleCrafter and DemoFusion in generating a 2048$\times$2048 image using a pre-trained SDXL model and achieves an FID$_b$ improvement of 11.6 and 3.7, respectively. FreCaS can be easily extended to more complex models such as SD3. The source code of FreCaS can be found at $\href{\text{https://github.com/xtudbxk/FreCaS}}{https://github.com/xtudbxk/FreCaS}$.
Diffusion models have recently gained recognition for generating diverse and high-quality content, especially in the domain of image synthesis. These models excel not only in creating fixed-size images but also in producing panoramic images. However, existing methods often struggle with spatial layout consistency when producing high-resolution panoramas, due to the lack of guidance of the global image layout. In this paper, we introduce the Multi-Scale Diffusion (MSD) framework, a plug-and-play module that extends the existing panoramic image generation framework to multiple resolution levels. By utilizing gradient descent techniques, our method effectively incorporates structural information from low-resolution images into high-resolution outputs. A comprehensive evaluation of the proposed method was conducted, comparing it with the prior works in qualitative and quantitative dimensions. The evaluation results demonstrate that our method significantly outperforms others in generating coherent high-resolution panoramas.
Recently, prompt learning has emerged as the state-of-the-art (SOTA) for fair text-to-image (T2I) generation. Specifically, this approach leverages readily available reference images to learn inclusive prompts for each target Sensitive Attribute (tSA), allowing for fair image generation. In this work, we first reveal that this prompt learning-based approach results in degraded sample quality. Our analysis shows that the approach's training objective -- which aims to align the embedding differences of learned prompts and reference images -- could be sub-optimal, resulting in distortion of the learned prompts and degraded generated images. To further substantiate this claim, as our major contribution, we deep dive into the denoising subnetwork of the T2I model to track down the effect of these learned prompts by analyzing the cross-attention maps. In our analysis, we propose a novel prompt switching analysis: I2H and H2I. Furthermore, we propose new quantitative characterization of cross-attention maps. Our analysis reveals abnormalities in the early denoising steps, perpetuating improper global structure that results in degradation in the generated samples. Building on insights from our analysis, we propose two ideas: (i) Prompt Queuing and (ii) Attention Amplification to address the quality issue. Extensive experimental results on a wide range of tSAs show that our proposed method outperforms SOTA approach's image generation quality, while achieving competitive fairness. More resources at FairQueue Project site: https://sutd-visual-computing-group.github.io/FairQueue
Large pre-trained models have achieved notable success across a range of downstream tasks. However, recent research shows that a type of adversarial attack ($\textit{i.e.,}$ backdoor attack) can manipulate the behavior of machine learning models through contaminating their training dataset, posing significant threat in the real-world application of large pre-trained model, especially for those customized models. Therefore, addressing the unique challenges for exploring vulnerability of pre-trained models is of paramount importance. Through empirical studies on the capability for performing backdoor attack in large pre-trained models ($\textit{e.g.,}$ ViT), we find the following unique challenges of attacking large pre-trained models: 1) the inability to manipulate or even access large training datasets, and 2) the substantial computational resources required for training or fine-tuning these models. To address these challenges, we establish new standards for an effective and feasible backdoor attack in the context of large pre-trained models. In line with these standards, we introduce our EDT model, an \textbf{E}fficient, \textbf{D}ata-free, \textbf{T}raining-free backdoor attack method. Inspired by model editing techniques, EDT injects an editing-based lightweight codebook into the backdoor of large pre-trained models, which replaces the embedding of the poisoned image with the target image without poisoning the training dataset or training the victim model. Our experiments, conducted across various pre-trained models such as ViT, CLIP, BLIP, and stable diffusion, and on downstream tasks including image classification, image captioning, and image generation, demonstrate the effectiveness of our method. Our code is available in the supplementary material.
Diffusion models achieve superior generation quality but suffer from slow generation speed due to the iterative nature of denoising. In contrast, consistency models, a new generative family, achieve competitive performance with significantly faster sampling. These models are trained either through consistency distillation, which leverages pretrained diffusion models, or consistency training/tuning directly from raw data. In this work, we propose a novel framework for understanding consistency models by modeling the denoising process of the diffusion model as a Markov Decision Process (MDP) and framing consistency model training as the value estimation through Temporal Difference~(TD) Learning. More importantly, this framework allows us to analyze the limitations of current consistency training/tuning strategies. Built upon Easy Consistency Tuning (ECT), we propose Stable Consistency Tuning (SCT), which incorporates variance-reduced learning using the score identity. SCT leads to significant performance improvements on benchmarks such as CIFAR-10 and ImageNet-64. On ImageNet-64, SCT achieves 1-step FID 2.42 and 2-step FID 1.55, a new SoTA for consistency models.
Visual text design plays a critical role in conveying themes, emotions, and atmospheres in multimodal formats such as film posters and album covers. Translating these visual and textual elements across languages extends the concept of translation beyond mere text, requiring the adaptation of aesthetic and stylistic features. To address this, we introduce a novel task of Multimodal Style Translation (MuST-Bench), a benchmark designed to evaluate the ability of visual text generation models to perform translation across different writing systems while preserving design intent. Our initial experiments on MuST-Bench reveal that existing visual text generation models struggle with the proposed task due to the inadequacy of textual descriptions in conveying visual design. In response, we introduce SIGIL, a framework for multimodal style translation that eliminates the need for style descriptions. SIGIL enhances image generation models through three innovations: glyph latent for multilingual settings, pretrained VAEs for stable style guidance, and an OCR model with reinforcement learning feedback for optimizing readable character generation. SIGIL outperforms existing baselines by achieving superior style consistency and legibility while maintaining visual fidelity, setting itself apart from traditional description-based approaches. We release MuST-Bench publicly for broader use and exploration https://huggingface.co/datasets/yejinc/MuST-Bench.
This paper introduces Ali-AUG, a novel single-step diffusion model for efficient labeled data augmentation in industrial applications. Our method addresses the challenge of limited labeled data by generating synthetic, labeled images with precise feature insertion. Ali-AUG utilizes a stable diffusion architecture enhanced with skip connections and LoRA modules to efficiently integrate masks and images, ensuring accurate feature placement without affecting unrelated image content. Experimental validation across various industrial datasets demonstrates Ali-AUG's superiority in generating high-quality, defect-enhanced images while maintaining rapid single-step inference. By offering precise control over feature insertion and minimizing required training steps, our technique significantly enhances data augmentation capabilities, providing a powerful tool for improving the performance of deep learning models in scenarios with limited labeled data. Ali-AUG is especially useful for use cases like defective product image generation to train AI-based models to improve their ability to detect defects in manufacturing processes. Using different data preparation strategies, including Classification Accuracy Score (CAS) and Naive Augmentation Score (NAS), we show that Ali-AUG improves model performance by 31% compared to other augmentation methods and by 45% compared to models without data augmentation. Notably, Ali-AUG reduces training time by 32% and supports both paired and unpaired datasets, enhancing flexibility in data preparation.
Regional defaults describe the emerging phenomenon that text-to-image (T2I) foundation models used in generative AI are prone to over-proportionally depicting certain geographic regions to the exclusion of others. In this work, we introduce a scalable evaluation for uncovering such regional defaults. The evaluation consists of region hierarchy--based image generation and cross-level similarity comparisons. We carry out an experiment by prompting DALL-E 2, a state-of-the-art T2I generation model capable of generating photorealistic images, to depict a forest. We select forest as an object class that displays regional variation and can be characterized using spatial statistics. For a region in the hierarchy, our experiment reveals the regional defaults implicit in DALL-E 2, along with their scale-dependent nature and spatial relationships. In addition, we discover that the implicit defaults do not necessarily correspond to the most widely forested regions in reality. Our findings underscore a need for further investigation into the geography of T2I generation and other forms of generative AI.
Text-to-Image (TTI) systems often support people during ideation, the early stages of a creative process when exposure to a broad set of relevant images can help explore the design space. Since ideation is an important subclass of TTI tasks, understanding how to quantitatively evaluate TTI systems according to how well they support ideation is crucial to promoting research and development for these users. However, existing evaluation metrics for TTI remain focused on distributional similarity metrics like Fr\'echet Inception Distance (FID). We take an alternative approach and, based on established methods from ranking evaluation, develop TTI evaluation metrics with explicit models of how users browse and interact with sets of spatially arranged generated images. Our proposed offline evaluation metrics for TTI not only capture how relevant generated images are with respect to the user's ideation need but also take into consideration the diversity and arrangement of the set of generated images. We analyze our proposed family of TTI metrics using human studies on image grids generated by three different TTI systems based on subsets of the widely used benchmarks such as MS-COCO captions and Localized Narratives as well as prompts used in naturalistic settings. Our results demonstrate that grounding metrics in how people use systems is an important and understudied area of benchmark design.
Existing reconstruction-based novel view synthesis methods for driving scenes focus on synthesizing camera views along the recorded trajectory of the ego vehicle. Their image rendering performance will severely degrade on viewpoints falling out of the recorded trajectory, where camera rays are untrained. We propose FreeVS, a novel fully generative approach that can synthesize camera views on free new trajectories in real driving scenes. To control the generation results to be 3D consistent with the real scenes and accurate in viewpoint pose, we propose the pseudo-image representation of view priors to control the generation process. Viewpoint transformation simulation is applied on pseudo-images to simulate camera movement in each direction. Once trained, FreeVS can be applied to any validation sequences without reconstruction process and synthesis views on novel trajectories. Moreover, we propose two new challenging benchmarks tailored to driving scenes, which are novel camera synthesis and novel trajectory synthesis, emphasizing the freedom of viewpoints. Given that no ground truth images are available on novel trajectories, we also propose to evaluate the consistency of images synthesized on novel trajectories with 3D perception models. Experiments on the Waymo Open Dataset show that FreeVS has a strong image synthesis performance on both the recorded trajectories and novel trajectories. Project Page: https://freevs24.github.io/
Direct Preference Optimization (DPO) has emerged as a powerful approach to align text-to-image (T2I) models with human feedback. Unfortunately, successful application of DPO to T2I models requires a huge amount of resources to collect and label large-scale datasets, e.g., millions of generated paired images annotated with human preferences. In addition, these human preference datasets can get outdated quickly as the rapid improvements of T2I models lead to higher quality images. In this work, we investigate a scalable approach for collecting large-scale and fully synthetic datasets for DPO training. Specifically, the preferences for paired images are generated using a pre-trained reward function, eliminating the need for involving humans in the annotation process, greatly improving the dataset collection efficiency. Moreover, we demonstrate that such datasets allow averaging predictions across multiple models and collecting ranked preferences as opposed to pairwise preferences. Furthermore, we introduce RankDPO to enhance DPO-based methods using the ranking feedback. Applying RankDPO on SDXL and SD3-Medium models with our synthetically generated preference dataset "Syn-Pic" improves both prompt-following (on benchmarks like T2I-Compbench, GenEval, and DPG-Bench) and visual quality (through user studies). This pipeline presents a practical and scalable solution to develop better preference datasets to enhance the performance of text-to-image models.
In recent years, diffusion models have emerged as a superior alternative to generative adversarial networks (GANs) for high-fidelity image generation, with wide applications in text-to-image generation, image-to-image translation, and super-resolution. However, their real-time feasibility is hindered by slow training and inference speeds. This study addresses this challenge by proposing a wavelet-based conditional Diffusion GAN scheme for Single-Image Super-Resolution (SISR). Our approach utilizes the diffusion GAN paradigm to reduce the timesteps required by the reverse diffusion process and the Discrete Wavelet Transform (DWT) to achieve dimensionality reduction, decreasing training and inference times significantly. The results of an experimental validation on the CelebA-HQ dataset confirm the effectiveness of our proposed scheme. Our approach outperforms other state-of-the-art methodologies successfully ensuring high-fidelity output while overcoming inherent drawbacks associated with diffusion models in time-sensitive applications.
The proliferation of machine learning models in diverse clinical applications has led to a growing need for high-fidelity, medical image training data. Such data is often scarce due to cost constraints and privacy concerns. Alleviating this burden, medical image synthesis via generative adversarial networks (GANs) emerged as a powerful method for synthetically generating photo-realistic images based on existing sets of real medical images. However, the exact image set size required to efficiently train such a GAN is unclear. In this work, we experimentally establish benchmarks that measure the relationship between a sample dataset size and the fidelity of the generated images, given the dataset's distribution of image complexities. We analyze statistical metrics based on delentropy, an image complexity measure rooted in Shannon's entropy in information theory. For our pipeline, we conduct experiments with two state-of-the-art GANs, StyleGAN 3 and SPADE-GAN, trained on multiple medical imaging datasets with variable sample sizes. Across both GANs, general performance improved with increasing training set size but suffered with increasing complexity.
Clinical decision-making relies heavily on causal reasoning and longitudinal analysis. For example, for a patient with Alzheimer's disease (AD), how will the brain grey matter atrophy in a year if intervened on the A-beta level in cerebrospinal fluid? The answer is fundamental to diagnosis and follow-up treatment. However, this kind of inquiry involves counterfactual medical images which can not be acquired by instrumental or correlation-based image synthesis models. Yet, such queries require counterfactual medical images, not obtainable through standard image synthesis models. Hence, a causal longitudinal image synthesis (CLIS) method, enabling the synthesis of such images, is highly valuable. However, building a CLIS model confronts three primary yet unmet challenges: mismatched dimensionality between high-dimensional images and low-dimensional tabular variables, inconsistent collection intervals of follow-up data, and inadequate causal modeling capability of existing causal graph methods for image data. In this paper, we established a tabular-visual causal graph (TVCG) for CLIS overcoming these challenges through a novel integration of generative imaging, continuous-time modeling, and structural causal models combined with a neural network. We train our CLIS based on the ADNI dataset and evaluate it on two other AD datasets, which illustrate the outstanding yet controllable quality of the synthesized images and the contributions of synthesized MRI to the characterization of AD progression, substantiating the reliability and utility in clinics.
Deep generative modeling has emerged as a powerful tool for synthesizing realistic medical images, driving advances in medical image analysis, disease diagnosis, and treatment planning. This chapter explores various deep generative models for 3D medical image synthesis, with a focus on Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), and Denoising Diffusion Models (DDMs). We discuss the fundamental principles, recent advances, as well as strengths and weaknesses of these models and examine their applications in clinically relevant problems, including unconditional and conditional generation tasks like image-to-image translation and image reconstruction. We additionally review commonly used evaluation metrics for assessing image fidelity, diversity, utility, and privacy and provide an overview of current challenges in the field.
Generative Adversarial Networks (GANs) have emerged as a prominent research focus for image editing tasks, leveraging the powerful image generation capabilities of the GAN framework to produce remarkable results.However, prevailing approaches are contingent upon extensive training datasets and explicit supervision, presenting a significant challenge in manipulating the diverse attributes of new image classes with limited sample availability. To surmount this hurdle, we introduce TAGE, an innovative image generation network comprising three integral modules: the Codebook Learning Module (CLM), the Code Prediction Module (CPM) and the Prompt-driven Semantic Module (PSM). The CPM module delves into the semantic dimensions of category-agnostic attributes, encapsulating them within a discrete codebook. This module is predicated on the concept that images are assemblages of attributes, and thus, by editing these category-independent attributes, it is theoretically possible to generate images from unseen categories. Subsequently, the CPM module facilitates naturalistic image editing by predicting indices of category-independent attribute vectors within the codebook. Additionally, the PSM module generates semantic cues that are seamlessly integrated into the Transformer architecture of the CPM, enhancing the model's comprehension of the targeted attributes for editing. With these semantic cues, the model can generate images that accentuate desired attributes more prominently while maintaining the integrity of the original category, even with a limited number of samples. We have conducted extensive experiments utilizing the Animal Faces, Flowers, and VGGFaces datasets. The results of these experiments demonstrate that our proposed method not only achieves superior performance but also exhibits a high degree of stability when compared to other few-shot image generation techniques.
Diffusion models have achieved remarkable progress in the field of image generation due to their outstanding capabilities. However, these models require substantial computing resources because of the multi-step denoising process during inference. While traditional pruning methods have been employed to optimize these models, the retraining process necessitates large-scale training datasets and extensive computational costs to maintain generalization ability, making it neither convenient nor efficient. Recent studies attempt to utilize the similarity of features across adjacent denoising stages to reduce computational costs through simple and static strategies. However, these strategies cannot fully harness the potential of the similar feature patterns across adjacent timesteps. In this work, we propose a novel pruning method that derives an efficient diffusion model via a more intelligent and differentiable pruner. At the core of our approach is casting the model pruning process into a SubNet search process. Specifically, we first introduce a SuperNet based on standard diffusion via adding some backup connections built upon the similar features. We then construct a plugin pruner network and design optimization losses to identify redundant computation. Finally, our method can identify an optimal SubNet through few-step gradient optimization and a simple post-processing procedure. We conduct extensive experiments on various diffusion models including Stable Diffusion series and DiTs. Our DiP-GO approach achieves 4.4 x speedup for SD-1.5 without any loss of accuracy, significantly outperforming the previous state-of-the-art methods.
Finding clusters of data points with similar characteristics and generating new cluster-specific samples can significantly enhance our understanding of complex data distributions. While clustering has been widely explored using Variational Autoencoders, these models often lack generation quality in real-world datasets. This paper addresses this gap by introducing TreeDiffusion, a deep generative model that conditions Diffusion Models on hierarchical clusters to obtain high-quality, cluster-specific generations. The proposed pipeline consists of two steps: a VAE-based clustering model that learns the hierarchical structure of the data, and a conditional diffusion model that generates realistic images for each cluster. We propose this two-stage process to ensure that the generated samples remain representative of their respective clusters and enhance image fidelity to the level of diffusion models. A key strength of our method is its ability to create images for each cluster, providing better visualization of the learned representations by the clustering model, as demonstrated through qualitative results. This method effectively addresses the generative limitations of VAE-based approaches while preserving their clustering performance. Empirically, we demonstrate that conditioning diffusion models on hierarchical clusters significantly enhances generative performance, thereby advancing the state of generative clustering models.
Movie posters are vital for captivating audiences, conveying themes, and driving market competition in the film industry. While traditional designs are laborious, intelligent generation technology offers efficiency gains and design enhancements. Despite exciting progress in image generation, current models often fall short in producing satisfactory poster results. The primary issue lies in the absence of specialized poster datasets for targeted model training. In this work, we propose a Movie Posters DataSet (MPDS), tailored for text-to-image generation models to revolutionize poster production. As dedicated to posters, MPDS stands out as the first image-text pair dataset to our knowledge, composing of 373k+ image-text pairs and 8k+ actor images (covering 4k+ actors). Detailed poster descriptions, such as movie titles, genres, casts, and synopses, are meticulously organized and standardized based on public movie synopsis, also named movie-synopsis prompt. To bolster poster descriptions as well as reduce differences from movie synopsis, further, we leverage a large-scale vision-language model to automatically produce vision-perceptive prompts for each poster, then perform manual rectification and integration with movie-synopsis prompt. In addition, we introduce a prompt of poster captions to exhibit text elements in posters like actor names and movie titles. For movie poster generation, we develop a multi-condition diffusion framework that takes poster prompt, poster caption, and actor image (for personalization) as inputs, yielding excellent results through the learning of a diffusion model. Experiments demonstrate the valuable role of our proposed MPDS dataset in advancing personalized movie poster generation. MPDS is available at https://anonymous.4open.science/r/MPDS-373k-BD3B.
Diffusion models have demonstrated remarkable capabilities in image synthesis, but their recently proven vulnerability to Membership Inference Attacks (MIAs) poses a critical privacy concern. This paper introduces two novel and efficient approaches (DualMD and DistillMD) to protect diffusion models against MIAs while maintaining high utility. Both methods are based on training two separate diffusion models on disjoint subsets of the original dataset. DualMD then employs a private inference pipeline that utilizes both models. This strategy significantly reduces the risk of black-box MIAs by limiting the information any single model contains about individual training samples. The dual models can also generate "soft targets" to train a private student model in DistillMD, enhancing privacy guarantees against all types of MIAs. Extensive evaluations of DualMD and DistillMD against state-of-the-art MIAs across various datasets in white-box and black-box settings demonstrate their effectiveness in substantially reducing MIA success rates while preserving competitive image generation performance. Notably, our experiments reveal that DistillMD not only defends against MIAs but also mitigates model memorization, indicating that both vulnerabilities stem from overfitting and can be addressed simultaneously with our unified approach.