We introduce OneDiffusion, a versatile, large-scale diffusion model that seamlessly supports bidirectional image synthesis and understanding across diverse tasks. It enables conditional generation from inputs such as text, depth, pose, layout, and semantic maps, while also handling tasks like image deblurring, upscaling, and reverse processes such as depth estimation and segmentation. Additionally, OneDiffusion allows for multi-view generation, camera pose estimation, and instant personalization using sequential image inputs. Our model takes a straightforward yet effective approach by treating all tasks as frame sequences with varying noise scales during training, allowing any frame to act as a conditioning image at inference time. Our unified training framework removes the need for specialized architectures, supports scalable multi-task training, and adapts smoothly to any resolution, enhancing both generalization and scalability. Experimental results demonstrate competitive performance across tasks in both generation and prediction such as text-to-image, multiview generation, ID preservation, depth estimation and camera pose estimation despite relatively small training dataset. Our code and checkpoint are freely available at https://github.com/lehduong/OneDiffusion
Multimodal MR image synthesis aims to generate missing modality images by effectively fusing and mapping from a subset of available MRI modalities. Most existing methods adopt an image-to-image translation paradigm, treating multiple modalities as input channels. However, these approaches often yield sub-optimal results due to the inherent difficulty in achieving precise feature- or semantic-level alignment across modalities. To address these challenges, we propose an Adaptive Group-wise Interaction Network (AGI-Net) that explicitly models both inter-modality and intra-modality relationships for multimodal MR image synthesis. Specifically, feature channels are first partitioned into predefined groups, after which an adaptive rolling mechanism is applied to conventional convolutional kernels to better capture feature and semantic correspondences between different modalities. In parallel, a cross-group attention module is introduced to enable effective feature fusion across groups, thereby enhancing the network's representational capacity. We validate the proposed AGI-Net on the publicly available IXI and BraTS2023 datasets. Experimental results demonstrate that AGI-Net achieves state-of-the-art performance in multimodal MR image synthesis tasks, confirming the effectiveness of its modality-aware interaction design. We release the relevant code at: https://github.com/zunzhumu/Adaptive-Group-wise-Interaction-Network-for-Multimodal-MRI-Synthesis.git.
Diffusion Transformers (DiTs) have exhibited robust capabilities in image generation tasks. However, accurate text-guided image editing for multimodal DiTs (MM-DiTs) still poses a significant challenge. Unlike UNet-based structures that could utilize self/cross-attention maps for semantic editing, MM-DiTs inherently lack support for explicit and consistent incorporated text guidance, resulting in semantic misalignment between the edited results and texts. In this study, we disclose the sensitivity of different attention heads to different image semantics within MM-DiTs and introduce HeadRouter, a training-free image editing framework that edits the source image by adaptively routing the text guidance to different attention heads in MM-DiTs. Furthermore, we present a dual-token refinement module to refine text/image token representations for precise semantic guidance and accurate region expression. Experimental results on multiple benchmarks demonstrate HeadRouter's performance in terms of editing fidelity and image quality.
Diffusion models have shown remarkable success in text-to-image generation, making alignment methods for these models increasingly important. A key challenge is the sparsity of preference labels, which are typically available only at the terminal of denoising trajectories. This raises the issue of how to assign credit across denoising steps based on these sparse labels. In this paper, we propose Denoised Distribution Estimation (DDE), a novel method for credit assignment. Unlike previous approaches that rely on auxiliary models or hand-crafted schemes, DDE derives its strategy more explicitly. The proposed DDE directly estimates the terminal denoised distribution from the perspective of each step. It is equipped with two estimation strategies and capable of representing the entire denoising trajectory with a single model inference. Theoretically and empirically, we show that DDE prioritizes optimizing the middle part of the denoising trajectory, resulting in a novel and effective credit assignment scheme. Extensive experiments demonstrate that our approach achieves superior performance, both quantitatively and qualitatively.
Diffusion models (DMs), which enable both image generation from noise and inversion from data, have inspired powerful unpaired image-to-image (I2I) translation algorithms. However, they often require a larger number of neural function evaluations (NFEs), limiting their practical applicability. In this paper, we tackle this problem with Schrodinger Bridges (SBs), which are stochastic differential equations (SDEs) between distributions with minimal transport cost. We analyze the probability flow ordinary differential equation (ODE) formulation of SBs, and observe that we can decompose its vector field into a linear combination of source predictor, target predictor, and noise predictor. Inspired by this observation, we propose Latent Schrodinger Bridges (LSBs) that approximate the SB ODE via pre-trained Stable Diffusion, and develop appropriate prompt optimization and change of variables formula to match the training and inference between distributions. We demonstrate that our algorithm successfully conduct competitive I2I translation in unsupervised setting with only a fraction of computation cost required by previous DM-based I2I methods.
Unmanned Aerial Vehicle (UAV) Cross-View Geo-Localization (CVGL) presents significant challenges due to the view discrepancy between oblique UAV images and overhead satellite images. Existing methods heavily rely on the supervision of labeled datasets to extract viewpoint-invariant features for cross-view retrieval. However, these methods have expensive training costs and tend to overfit the region-specific cues, showing limited generalizability to new regions. To overcome this issue, we propose an unsupervised solution that lifts the scene representation to 3d space from UAV observations for satellite image generation, providing robust representation against view distortion. By generating orthogonal images that closely resemble satellite views, our method reduces view discrepancies in feature representation and mitigates shortcuts in region-specific image pairing. To further align the rendered image's perspective with the real one, we design an iterative camera pose updating mechanism that progressively modulates the rendered query image with potential satellite targets, eliminating spatial offsets relative to the reference images. Additionally, this iterative refinement strategy enhances cross-view feature invariance through view-consistent fusion across iterations. As such, our unsupervised paradigm naturally avoids the problem of region-specific overfitting, enabling generic CVGL for UAV images without feature fine-tuning or data-driven training. Experiments on the University-1652 and SUES-200 datasets demonstrate that our approach significantly improves geo-localization accuracy while maintaining robustness across diverse regions. Notably, without model fine-tuning or paired training, our method achieves competitive performance with recent supervised methods.
Denoising with a Joint-Embedding Predictive Architecture (D-JEPA), an autoregressive model, has demonstrated outstanding performance in class-conditional image generation. However, the application of next-token prediction in high-resolution text-to-image generation remains underexplored. In this paper, we introduce D-JEPA$\cdot$T2I, an extension of D-JEPA incorporating flow matching loss, designed to enable data-efficient continuous resolution learning. D-JEPA$\cdot$T2I leverages a multimodal visual transformer to effectively integrate textual and visual features and adopts Visual Rotary Positional Embedding (VoPE) to facilitate continuous resolution learning. Furthermore, we devise a data feedback mechanism that significantly enhances data utilization efficiency. For the first time, we achieve state-of-the-art \textbf{high-resolution} image synthesis via next-token prediction. The experimental code and pretrained models will be open-sourced at \url{https://d-jepa.github.io/t2i}.
Significant advancements have been made in semantic image synthesis in remote sensing. However, existing methods still face formidable challenges in balancing semantic controllability and diversity. In this paper, we present a Hybrid Semantic Embedding Guided Generative Adversarial Network (HySEGGAN) for controllable and efficient remote sensing image synthesis. Specifically, HySEGGAN leverages hierarchical information from a single source. Motivated by feature description, we propose a hybrid semantic Embedding method, that coordinates fine-grained local semantic layouts to characterize the geometric structure of remote sensing objects without extra information. Besides, a Semantic Refinement Network (SRN) is introduced, incorporating a novel loss function to ensure fine-grained semantic feedback. The proposed approach mitigates semantic confusion and prevents geometric pattern collapse. Experimental results indicate that the method strikes an excellent balance between semantic controllability and diversity. Furthermore, HySEGGAN significantly improves the quality of synthesized images and achieves state-of-the-art performance as a data augmentation technique across multiple datasets for downstream tasks.
As text-to-image models grow increasingly powerful and complex, their burgeoning size presents a significant obstacle to widespread adoption, especially on resource-constrained devices. This paper presents a pioneering study on post-training pruning of Stable Diffusion 2, addressing the critical need for model compression in text-to-image domain. Our study tackles the pruning techniques for the previously unexplored multi-modal generation models, and particularly examines the pruning impact on the textual component and the image generation component separately. We conduct a comprehensive comparison on pruning the model or the single component of the model in various sparsities. Our results yield previously undocumented findings. For example, contrary to established trends in language model pruning, we discover that simple magnitude pruning outperforms more advanced techniques in text-to-image context. Furthermore, our results show that Stable Diffusion 2 can be pruned to 38.5% sparsity with minimal quality loss, achieving a significant reduction in model size. We propose an optimal pruning configuration that prunes the text encoder to 47.5% and the diffusion generator to 35%. This configuration maintains image generation quality while substantially reducing computational requirements. In addition, our work uncovers intriguing questions about information encoding in text-to-image models: we observe that pruning beyond certain thresholds leads to sudden performance drops (unreadable images), suggesting that specific weights encode critical semantics information. This finding opens new avenues for future research in model compression, interoperability, and bias identification in text-to-image models. By providing crucial insights into the pruning behavior of text-to-image models, our study lays the groundwork for developing more efficient and accessible AI-driven image generation systems
Due to the singularity of real-world paired datasets and the complexity of low-light environments, this leads to supervised methods lacking a degree of scene generalisation. Meanwhile, limited by poor lighting and content guidance, existing zero-shot methods cannot handle unknown severe degradation well. To address this problem, we will propose a new zero-shot low-light enhancement method to compensate for the lack of light and structural information in the diffusion sampling process by effectively combining the wavelet and Fourier frequency domains to construct rich a priori information. The key to the inspiration comes from the similarity between the wavelet and Fourier frequency domains: both light and structure information are closely related to specific frequency domain regions, respectively. Therefore, by transferring the diffusion process to the wavelet low-frequency domain and combining the wavelet and Fourier frequency domains by continuously decomposing them in the inverse process, the constructed rich illumination prior is utilised to guide the image generation enhancement process. Sufficient experiments show that the framework is robust and effective in various scenarios. The code will be available at: \href{https://github.com/hejh8/Joint-Wavelet-and-Fourier-priors-guided-diffusion}{https://github.com/hejh8/Joint-Wavelet-and-Fourier-priors-guided-diffusion}.
Image generation has shown remarkable results in generating high-fidelity realistic images, in particular with the advancement of diffusion-based models. However, the prevalence of AI-generated images may have side effects for the machine learning community that are not clearly identified. Meanwhile, the success of deep learning in computer vision is driven by the massive dataset collected on the Internet. The extensive quantity of synthetic data being added to the Internet would become an obstacle for future researchers to collect "clean" datasets without AI-generated content. Prior research has shown that using datasets contaminated by synthetic images may result in performance degradation when used for training. In this paper, we investigate the potential impact of contaminated datasets on Online Continual Learning (CL) research. We experimentally show that contaminated datasets might hinder the training of existing online CL methods. Also, we propose Entropy Selection with Real-synthetic similarity Maximization (ESRM), a method to alleviate the performance deterioration caused by synthetic images when training online CL models. Experiments show that our method can significantly alleviate performance deterioration, especially when the contamination is severe. For reproducibility, the source code of our work is available at https://github.com/maorong-wang/ESRM.
Despite recent advancements, text-to-image generation models often produce images containing artifacts, especially in human figures. These artifacts appear as poorly generated human bodies, including distorted, missing, or extra body parts, leading to visual inconsistencies with typical human anatomy and greatly impairing overall fidelity. In this study, we address this challenge by curating Human Artifact Dataset (HAD), the first large-scale dataset specifically designed to identify and localize human artifacts. HAD comprises over 37,000 images generated by several popular text-to-image models, annotated for human artifact localization. Using this dataset, we train the Human Artifact Detection Models (HADM), which can identify diverse artifact types across multiple generative domains and demonstrate strong generalization, even on images from unseen generators. Additionally, to further improve generators' perception of human structural coherence, we use the predictions from our HADM as feedback for diffusion model finetuning. Our experiments confirm a reduction in human artifacts in the resulting model. Furthermore, we showcase a novel application of our HADM in an iterative inpainting framework to correct human artifacts in arbitrary images directly, demonstrating its utility in improving image quality. Our dataset and detection models are available at: \url{https://github.com/wangkaihong/HADM}.
Training of large-scale text-to-image and image-to-image models requires a huge amount of annotated data. While text-to-image datasets are abundant, data available for instruction-based image-to-image tasks like object addition and removal is limited. This is because of the several challenges associated with the data generation process, such as, significant human effort, limited automation, suboptimal end-to-end models, data diversity constraints and high expenses. We propose an automated data generation pipeline aimed at alleviating such limitations, and introduce GalaxyEdit - a large-scale image editing dataset for add and remove operations. We fine-tune the SD v1.5 model on our dataset and find that our model can successfully handle a broader range of objects and complex editing instructions, outperforming state-of-the-art methods in FID scores by 11.2\% and 26.1\% for add and remove tasks respectively. Furthermore, in light of on-device usage scenarios, we expand our research to include task-specific lightweight adapters leveraging the ControlNet-xs architecture. While ControlNet-xs excels in canny and depth guided generation, we propose to improve the communication between the control network and U-Net for more intricate add and remove tasks. We achieve this by enhancing ControlNet-xs with non-linear interaction layers based on Volterra filters. Our approach outperforms ControlNet-xs in both add/remove and canny-guided image generation tasks, highlighting the effectiveness of the proposed enhancement.
Large text-to-image models demonstrate impressive generation capabilities; however, their substantial size necessitates expensive cloud servers for deployment. Conversely, light-weight models can be deployed on edge devices at lower cost but often with inferior generation quality for complex user prompts. To strike a balance between performance and cost, we propose a routing framework, called \texttt{RouteT2I}, which dynamically selects either the large cloud model or the light-weight edge model for each user prompt. Since generated image quality is challenging to measure directly, \texttt{RouteT2I} establishes multi-dimensional quality metrics, particularly, by evaluating the similarity between the generated images and both positive and negative texts that describe each specific quality metric. \texttt{RouteT2I} then predicts the expected quality of the generated images by identifying key tokens in the prompt and comparing their impact on the quality. \texttt{RouteT2I} further introduces the Pareto relative superiority to compare the multi-metric quality of the generated images. Based on this comparison and predefined cost constraints, \texttt{RouteT2I} allocates prompts to either the edge or the cloud. Evaluation reveals that \texttt{RouteT2I} significantly reduces the number of requesting large cloud model while maintaining high-quality image generation.
Automatic image generation is no longer just of interest to researchers, but also to practitioners. However, current models are sensitive to the settings used and automatic optimization methods often require human involvement. To bridge this gap, we introduce ComfyGI, a novel approach to automatically improve workflows for image generation without the need for human intervention driven by techniques from genetic improvement. This enables image generation with significantly higher quality in terms of the alignment with the given description and the perceived aesthetics. On the performance side, we find that overall, the images generated with an optimized workflow are about 50% better compared to the initial workflow in terms of the median ImageReward score. These already good results are even surpassed in our human evaluation, as the participants preferred the images improved by ComfyGI in around 90% of the cases.
Large Multimodal Models (LMMs) have demonstrated remarkable capabilities. While existing benchmarks for evaluating LMMs mainly focus on image comprehension, few works evaluate them from the image generation perspective. To address this issue, we propose a straightforward automated evaluation pipeline. Specifically, this pipeline requires LMMs to generate an image-prompt from a given input image. Subsequently, it employs text-to-image generative models to create a new image based on these generated prompts. Finally, we evaluate the performance of LMMs by comparing the original image with the generated one. Furthermore, we introduce MMGenBench-Test, a comprehensive benchmark developed to evaluate LMMs across 13 distinct image patterns, and MMGenBench-Domain, targeting the performance evaluation of LMMs within the generative image domain. A thorough evaluation involving over 50 popular LMMs demonstrates the effectiveness and reliability in both the pipeline and benchmark. Our observations indicate that numerous LMMs excelling in existing benchmarks fail to adequately complete the basic tasks, related to image understanding and description. This finding highlights the substantial potential for performance improvement in current LMMs and suggests avenues for future model optimization. Concurrently, our pipeline facilitates the efficient assessment of LMMs performance across diverse domains by using solely image inputs.
Training multimodal generative models on large, uncurated datasets can result in users being exposed to harmful, unsafe and controversial or culturally-inappropriate outputs. While model editing has been proposed to remove or filter undesirable concepts in embedding and latent spaces, it can inadvertently damage learned manifolds, distorting concepts in close semantic proximity. We identify limitations in current model editing techniques, showing that even benign, proximal concepts may become misaligned. To address the need for safe content generation, we propose a modular, dynamic solution that leverages safety-context embeddings and a dual reconstruction process using tunable weighted summation in the latent space to generate safer images. Our method preserves global context without compromising the structural integrity of the learned manifolds. We achieve state-of-the-art results on safe image generation benchmarks, while offering controllable variation of model safety. We identify trade-offs between safety and censorship, which presents a necessary perspective in the development of ethical AI models. We will release our code. Keywords: Text-to-Image Models, Generative AI, Safety, Reliability, Model Editing
Recovering the useful signal from seismic field data is critical in seismic data processing. Seismic field data are usually coupled by a useful signal and field noise (random noise with unknown distribution), making them difficult to decouple. Unfortunately, subject to the assumption biases of data prior and noise prior, different random noise attenuation methods may have different performance biases. Suppose data-driven supervised deep learning methods have plenty of labeled [real noisy (field) and clean (useful)] data pairs. In that case, they will learn useful information from the labeled dataset and relax the biases of the data-prior and noise-prior assumptions. To this end, we first use the invertible neural network (INN) to disentangle the field data in observational space into the latent variable in latent space. Then, by manipulating the latent variable’s partitions encoding high- and low-frequency information, INN can generate quality-controlled fake field data and decouple useful signal and field noise parts from field data in its backward pass. To gain decoupling and generative capabilities, the training of our INN only requires a relatively small labeled dataset containing field–useful data pairs. Sampling in latent space, the trained INN can generate an infinite number of paired (fake field and useful) samples. Experiments show that our method can effectively decouple useful signal and field noise, and the noise of the fake field data is close to field noise. The generated paired dataset can benefit downstream tasks, such as field noise attenuation.
Generative Adversarial Networks (GANs), particularly StyleGAN and its variants, have demonstrated remarkable capabilities in generating highly realistic images. Despite their success, adapting these models to diverse tasks such as domain adaptation, reference-guided synthesis, and text-guided manipulation with limited training data remains challenging. Towards this end, in this study, we present a novel framework that significantly extends the capabilities of a pre-trained StyleGAN by integrating CLIP space via hypernetworks. This integration allows dynamic adaptation of StyleGAN to new domains defined by reference images or textual descriptions. Additionally, we introduce a CLIP-guided discriminator that enhances the alignment between generated images and target domains, ensuring superior image quality. Our approach demonstrates unprecedented flexibility, enabling text-guided image manipulation without the need for text-specific training data and facilitating seamless style transfer. Comprehensive qualitative and quantitative evaluations confirm the robustness and superior performance of our framework compared to existing methods.
Diffusion models excel in image generation, but controlling them remains a challenge. We focus on the problem of style-conditioned image generation. Although example images work, they are cumbersome: srefs (style-reference codes) from MidJourney solve this issue by expressing a specific image style in a short numeric code. These have seen widespread adoption throughout social media due to both their ease of sharing and the fact they allow using an image for style control, without having to post the source images themselves. However, users are not able to generate srefs from their own images, nor is the underlying training procedure public. We propose StyleCodes: an open-source and open-research style encoder architecture and training procedure to express image style as a 20-symbol base64 code. Our experiments show that our encoding results in minimal loss in quality compared to traditional image-to-style techniques.