In computer vision, it is well-known that a lack of data diversity will impair model performance. In this study, we address the challenges of enhancing the dataset diversity problem in order to benefit various downstream tasks such as object detection and instance segmentation. We propose a simple yet effective data augmentation approach by leveraging advancements in generative models, specifically text-to-image synthesis technologies like Stable Diffusion. Our method focuses on generating variations of labeled real images, utilizing generative object and background augmentation via inpainting to augment existing training data without the need for additional annotations. We find that background augmentation, in particular, significantly improves the models' robustness and generalization capabilities. We also investigate how to adjust the prompt and mask to ensure the generated content comply with the existing annotations. The efficacy of our augmentation techniques is validated through comprehensive evaluations of the COCO dataset and several other key object detection benchmarks, demonstrating notable enhancements in model performance across diverse scenarios. This approach offers a promising solution to the challenges of dataset enhancement, contributing to the development of more accurate and robust computer vision models.
Accurate text segmentation results are crucial for text-related generative tasks, such as text image generation, text editing, text removal, and text style transfer. Recently, some scene text segmentation methods have made significant progress in segmenting regular text. However, these methods perform poorly in scenarios containing artistic text. Therefore, this paper focuses on the more challenging task of artistic text segmentation and constructs a real artistic text segmentation dataset. One challenge of the task is that the local stroke shapes of artistic text are changeable with diversity and complexity. We propose a decoder with the layer-wise momentum query to prevent the model from ignoring stroke regions of special shapes. Another challenge is the complexity of the global topological structure. We further design a skeleton-assisted head to guide the model to focus on the global structure. Additionally, to enhance the generalization performance of the text segmentation model, we propose a strategy for training data synthesis, based on the large multi-modal model and the diffusion model. Experimental results show that our proposed method and synthetic dataset can significantly enhance the performance of artistic text segmentation and achieve state-of-the-art results on other public datasets.
Recent breakthroughs in diffusion models have exhibited exceptional image-generation capabilities. However, studies show that some outputs are merely replications of training data. Such replications present potential legal challenges for model owners, especially when the generated content contains proprietary information. In this work, we introduce a straightforward yet effective method for detecting memorized prompts by inspecting the magnitude of text-conditional predictions. Our proposed method seamlessly integrates without disrupting sampling algorithms, and delivers high accuracy even at the first generation step, with a single generation per prompt. Building on our detection strategy, we unveil an explainable approach that shows the contribution of individual words or tokens to memorization. This offers an interactive medium for users to adjust their prompts. Moreover, we propose two strategies i.e., to mitigate memorization by leveraging the magnitude of text-conditional predictions, either through minimization during inference or filtering during training. These proposed strategies effectively counteract memorization while maintaining high-generation quality. Code is available at https://github.com/YuxinWenRick/diffusion_memorization.
Incorporating a temporal dimension into pretrained image diffusion models for video generation is a prevalent approach. However, this method is computationally demanding and necessitates large-scale video datasets. More critically, the heterogeneity between image and video datasets often results in catastrophic forgetting of the image expertise. Recent attempts to directly extract video snippets from image diffusion models have somewhat mitigated these problems. Nevertheless, these methods can only generate brief video clips with simple movements and fail to capture fine-grained motion or non-grid deformation. In this paper, we propose a novel Zero-Shot video Sampling algorithm, denoted as $\mathcal{ZS}^2$, capable of directly sampling high-quality video clips from existing image synthesis methods, such as Stable Diffusion, without any training or optimization. Specifically, $\mathcal{ZS}^2$ utilizes the dependency noise model and temporal momentum attention to ensure content consistency and animation coherence, respectively. This ability enables it to excel in related tasks, such as conditional and context-specialized video generation and instruction-guided video editing. Experimental results demonstrate that $\mathcal{ZS}^2$ achieves state-of-the-art performance in zero-shot video generation, occasionally outperforming recent supervised methods. Homepage: \url{https://densechen.github.io/zss/}.
The Gaussian diffusion model, initially designed for image generation, has recently been adapted for 3D point cloud generation. However, these adaptations have not fully considered the intrinsic geometric characteristics of 3D shapes, thereby constraining the diffusion model's potential for 3D shape manipulation. To address this limitation, we introduce a novel deformable 3D shape diffusion model that facilitates comprehensive 3D shape manipulation, including point cloud generation, mesh deformation, and facial animation. Our approach innovatively incorporates a differential deformation kernel, which deconstructs the generation of geometric structures into successive non-rigid deformation stages. By leveraging a probabilistic diffusion model to simulate this step-by-step process, our method provides a versatile and efficient solution for a wide range of applications, spanning from graphics rendering to facial expression animation. Empirical evidence highlights the effectiveness of our approach, demonstrating state-of-the-art performance in point cloud generation and competitive results in mesh deformation. Additionally, extensive visual demonstrations reveal the significant potential of our approach for practical applications. Our method presents a unique pathway for advancing 3D shape manipulation and unlocking new opportunities in the realm of virtual reality.
Recent advancements in image synthesis, particularly with the advent of GAN and Diffusion models, have amplified public concerns regarding the dissemination of disinformation. To address such concerns, numerous AI-generated Image (AIGI) Detectors have been proposed and achieved promising performance in identifying fake images. However, there still lacks a systematic understanding of the adversarial robustness of AIGI detectors. In this paper, we examine the vulnerability of state-of-the-art AIGI detectors against adversarial attack under white-box and black-box settings, which has been rarely investigated so far. To this end, we propose a new method to attack AIGI detectors. First, inspired by the obvious difference between real images and fake images in the frequency domain, we add perturbations under the frequency domain to push the image away from its original frequency distribution. Second, we explore the full posterior distribution of the surrogate model to further narrow this gap between heterogeneous AIGI detectors, e.g. transferring adversarial examples across CNNs and ViTs. This is achieved by introducing a novel post-train Bayesian strategy that turns a single surrogate into a Bayesian one, capable of simulating diverse victim models using one pre-trained surrogate, without the need for re-training. We name our method as Frequency-based Post-train Bayesian Attack, or FPBA. Through FPBA, we show that adversarial attack is truly a real threat to AIGI detectors, because FPBA can deliver successful black-box attacks across models, generators, defense methods, and even evade cross-generator detection, which is a crucial real-world detection scenario. The code will be shared upon acceptance.
This paper introduces a novel approach to illumination manipulation in diffusion models, addressing the gap in conditional image generation with a focus on lighting conditions. We conceptualize the diffusion model as a black-box image render and strategically decompose its energy function in alignment with the image formation model. Our method effectively separates and controls illumination-related properties during the generative process. It generates images with realistic illumination effects, including cast shadow, soft shadow, and inter-reflections. Remarkably, it achieves this without the necessity for learning intrinsic decomposition, finding directions in latent space, or undergoing additional training with new datasets.
While most vision tasks are essentially visual in nature (for recognition), some important tasks, especially in the medical field, also require quantitative analysis (for quantification) using quantitative images. Unlike in visual analysis, pixel values in quantitative images correspond to physical metrics measured by specific devices (e.g., a depth image). However, recent work has shown that it is sometimes possible to synthesize accurate quantitative values from visual ones (e.g., depth from visual cues or defocus). This research aims to improve quantitative image synthesis (QIS) by exploring pretraining and image resolution scaling. We propose a benchmark for evaluating pretraining performance using the task of QIS-based bone mineral density (BMD) estimation from plain X-ray images, where the synthesized quantitative image is used to derive BMD. Our results show that appropriate pretraining can improve QIS performance, significantly raising the correlation of BMD estimation from 0.820 to 0.898, while others do not help or even hinder it. Scaling-up the resolution can further boost the correlation up to 0.923, a significant enhancement over conventional methods. Future work will include exploring more pretraining strategies and validating them on other image synthesis tasks.
Usually, inverse problems are ill-posed. The solution to the inverse problem is indeterminate, meaning that for given observational data, there may be multiple possible solutions. It is not sufficient to give a definite solution to common seismic inverse problems. In this study, we provide stochastic solutions for seismic inverse problems (denoising and reconstruction). We sample a range of possible and high -quality solutions for a given observation with various degradations from the posterior distribution through Langevin dynamics with conditional score function, all shown to be reasonable results; for example, the stochastic solutions we sampled may contain as many geological structures of interest to the expert as possible. Experimental results on synthetic and field data verify the superiority of posterior sampling. In particular, our method has obvious advantages over other methods, such as traditional and (supervised, self-supervised, and unsupervised) deep learning (DL) methods, especially in denoising under extremely low signal-to-noise ratio (SNR) and reconstruction for data with consecutively missing traces and noise. We also analyze the advantages of our approach and concluded that successful generative modeling of seismic data by the score-based generative models (SGMs) is the key to posterior sampling for the inverse problems, which all benefit from the seismic data prior implicit in the trained score network in the SGMs.
The availability of seismic data has specific implications for inversion and interpretation methods, especially for learning-based methods such as deep learning. Forward modeling or acquisition of field data are common means of obtaining seismic data, which can be costly, time-consuming and labor-intensive. This paper provides a generative modeling method for generating seismic data. We use a score-based generative model to learn the distribution of target seismic data set. Specifically, we learn the gradient of the target data set distribution through score matching using the noise conditional score network (NCSN). Then, we sample high-quality samples similar to the target data set through Langevin dynamics with learned NCSN. We take the seismic records synthesized by Marmousi as an example to show the powerful generative modeling capabilities of the generative model. By sampling from a prior distribution (Gaussian distribution), the generative model can generate diverse samples and have good interpretability. For example, by interpolating two random data points from the prior distribution, the generated data has manifold continuity in certain features (such as amplitude, inclination, polarity, number of events, etc.).
Intelligent instruction-following robots capable of improving from autonomously collected experience have the potential to transform robot learning: instead of collecting costly teleoperated demonstration data, large-scale deployment of fleets of robots can quickly collect larger quantities of autonomous data that can collectively improve their performance. However, autonomous improvement requires solving two key problems: (i) fully automating a scalable data collection procedure that can collect diverse and semantically meaningful robot data and (ii) learning from non-optimal, autonomous data with no human annotations. To this end, we propose a novel approach that addresses these challenges, allowing instruction-following policies to improve from autonomously collected data without human supervision. Our framework leverages vision-language models to collect and evaluate semantically meaningful experiences in new environments, and then utilizes a decomposition of instruction following tasks into (semantic) language-conditioned image generation and (non-semantic) goal reaching, which makes it significantly more practical to improve from this autonomously collected data without any human annotations. We carry out extensive experiments in the real world to demonstrate the effectiveness of our approach, and find that in a suite of unseen environments, the robot policy can be improved 2x with autonomously collected data. We open-source the code for our semantic autonomous improvement pipeline, as well as our autonomous dataset of 30.5K trajectories collected across five tabletop environments.
The Diffusion models, widely used for image generation, face significant challenges related to their broad applicability due to prolonged inference times and high memory demands. Efficient Post-Training Quantization (PTQ) is crucial to address these issues. However, unlike traditional models, diffusion models critically rely on the time-step for the multi-round denoising. Typically, each time-step is encoded into a hypersensitive temporal feature by several modules. Despite this, existing PTQ methods do not optimize these modules individually. Instead, they employ unsuitable reconstruction objectives and complex calibration methods, leading to significant disturbances in the temporal feature and denoising trajectory, as well as reduced compression efficiency. To address these challenges, we introduce a novel quantization framework that includes three strategies: 1) TIB-based Maintenance: Based on our innovative Temporal Information Block (TIB) definition, Temporal Information-aware Reconstruction (TIAR) and Finite Set Calibration (FSC) are developed to efficiently align original temporal features. 2) Cache-based Maintenance: Instead of indirect and complex optimization for the related modules, pre-computing and caching quantized counterparts of temporal features are developed to minimize errors. 3) Disturbance-aware Selection: Employ temporal feature errors to guide a fine-grained selection between the two maintenance strategies for further disturbance reduction. This framework preserves most of the temporal information and ensures high-quality end-to-end generation. Extensive testing on various datasets, diffusion models and hardware confirms our superior performance and acceleration..
This work aims to address the multi-view perspective RGB generation from text prompts given Bird-Eye-View(BEV) semantics. Unlike prior methods that neglect layout consistency, lack the ability to handle detailed text prompts, or are incapable of generalizing to unseen view points, MVPbev simultaneously generates cross-view consistent images of different perspective views with a two-stage design, allowing object-level control and novel view generation at test-time. Specifically, MVPbev firstly projects given BEV semantics to perspective view with camera parameters, empowering the model to generalize to unseen view points. Then we introduce a multi-view attention module where special initialization and de-noising processes are introduced to explicitly enforce local consistency among overlapping views w.r.t. cross-view homography. Last but not least, MVPbev further allows test-time instance-level controllability by refining a pre-trained text-to-image diffusion model. Our extensive experiments on NuScenes demonstrate that our method is capable of generating high-resolution photorealistic images from text descriptions with thousands of training samples, surpassing the state-of-the-art methods under various evaluation metrics. We further demonstrate the advances of our method in terms of generalizability and controllability with the help of novel evaluation metrics and comprehensive human analysis. Our code, data, and model can be found in \url{https://github.com/kkaiwwana/MVPbev}.
We present Perm, a learned parametric representation of human 3D hair designed to facilitate various hair-related applications. Unlike previous work that jointly models the global hair structure and local curl patterns, we propose to disentangle them using a PCA-based strand representation in the frequency domain, thereby allowing more precise editing and output control. Specifically, we leverage our strand representation to fit and decompose hair geometry textures into low- to high-frequency hair structures, termed guide textures and residual textures, respectively. These decomposed textures are later parameterized with different generative models, emulating common stages in the hair grooming process. We conduct extensive experiments to validate the architecture design of Perm, and finally deploy the trained model as a generic prior to solve task-agnostic problems, further showcasing its flexibility and superiority in tasks such as single-view hair reconstruction, hairstyle editing, and hair-conditioned image generation. More details can be found on our project page: https://cs.yale.edu/homes/che/projects/perm/.
Vision-language foundation models such as CLIP have achieved tremendous results in global vision-language alignment, but still show some limitations in creating representations for specific image regions. % To address this problem, we propose MaskInversion, a method that leverages the feature representations of pre-trained foundation models, such as CLIP, to generate a context-aware embedding for a query image region specified by a mask at test time. MaskInversion starts with initializing an embedding token and compares its explainability map, derived from the foundation model, to the query mask. The embedding token is then subsequently refined to approximate the query region by minimizing the discrepancy between its explainability map and the query mask. During this process, only the embedding vector is updated, while the underlying foundation model is kept frozen allowing to use MaskInversion with any pre-trained model. As deriving the explainability map involves computing its gradient, which can be expensive, we propose a gradient decomposition strategy that simplifies this computation. The learned region representation can be used for a broad range of tasks, including open-vocabulary class retrieval, referring expression comprehension, as well as for localized captioning and image generation. We evaluate the proposed method on all those tasks on several datasets such as PascalVOC, MSCOCO, RefCOCO, and OpenImagesV7 and show its capabilities compared to other SOTA approaches.
Text-to-image generative models often present issues regarding fairness with respect to certain sensitive attributes, such as gender or skin tone. This study aims to reproduce the results presented in "ITI-GEN: Inclusive Text-to-Image Generation" by Zhang et al. (2023a), which introduces a model to improve inclusiveness in these kinds of models. We show that most of the claims made by the authors about ITI-GEN hold: it improves the diversity and quality of generated images, it is scalable to different domains, it has plug-and-play capabilities, and it is efficient from a computational point of view. However, ITI-GEN sometimes uses undesired attributes as proxy features and it is unable to disentangle some pairs of (correlated) attributes such as gender and baldness. In addition, when the number of considered attributes increases, the training time grows exponentially and ITI-GEN struggles to generate inclusive images for all elements in the joint distribution. To solve these issues, we propose using Hard Prompt Search with negative prompting, a method that does not require training and that handles negation better than vanilla Hard Prompt Search. Nonetheless, Hard Prompt Search (with or without negative prompting) cannot be used for continuous attributes that are hard to express in natural language, an area where ITI-GEN excels as it is guided by images during training. Finally, we propose combining ITI-GEN and Hard Prompt Search with negative prompting.
With the rapid development of Text-to-Image (T2I) models, biases in human image generation against demographic social groups become a significant concern, impacting fairness and ethical standards in AI. Some researchers propose their methods to tackle with the issue. However, existing methods are designed for specific models with fixed prompts, limiting their adaptability to the fast-evolving models and diverse practical scenarios. Moreover, they neglect the impact of hallucinations, leading to discrepancies between expected and actual results. To address these issues, we introduce VersusDebias, a novel and universal debiasing framework for biases in arbitrary T2I models, consisting of an array generation (AG) module and an image generation (IG) module. The self-adaptive AG module generates specialized attribute arrays to post-process hallucinations and debias multiple attributes simultaneously. The IG module employs a small language model to modify prompts according to the arrays and drives the T2I model to generate debiased images, enabling zero-shot debiasing. Extensive experiments demonstrate VersusDebias's capability to debias any models across gender, race, and age simultaneously. In both zero-shot and few-shot scenarios, VersusDebias outperforms existing methods, showcasing its exceptional utility. Our work is accessible at https://github.com/VersusDebias/VersusDebias to ensure reproducibility and facilitate further research.
The success of generative models has reached a unique threshold where their outputs are indistinguishable from real data, leading to the inevitable contamination of future data collection pipelines with synthetic data. While their potential to generate infinite samples initially offers promise for reducing data collection costs and addressing challenges in data-scarce fields, the severe degradation in performance has been observed when iterative loops of training and generation occur -- known as ``model collapse.'' This paper explores a practical scenario in which a pretrained text-to-image diffusion model is finetuned using synthetic images generated from a previous iteration, a process we refer to as the ``Chain of Diffusion.'' We first demonstrate the significant degradation in image quality caused by this iterative process and identify the key factor driving this decline through rigorous empirical investigations. Drawing an analogy between the Chain of Diffusion and biological evolution, we then introduce a novel theoretical analysis based on quantitative trait modeling. Our theoretical analysis aligns with empirical observations of the generated images in the Chain of Diffusion. Finally, we propose Reusable Diffusion Finetuning (ReDiFine), a simple yet effective strategy inspired by genetic mutations. ReDiFine mitigates model collapse without requiring any hyperparameter tuning, making it a plug-and-play solution for reusable image generation.
Immunofluorescent (IF) imaging is crucial for visualizing biomarker expressions, cell morphology and assessing the effects of drug treatments on sub-cellular components. IF imaging needs extra staining process and often requiring cell fixation, therefore it may also introduce artefects and alter endogenouous cell morphology. Some IF stains are expensive or not readily available hence hindering experiments. Recent diffusion models, which synthesise high-fidelity IF images from easy-to-acquire brightfield (BF) images, offer a promising solution but are hindered by training instability and slow inference times due to the noise diffusion process. This paper presents a novel method for the conditional synthesis of IF images directly from BF images along with cell segmentation masks. Our approach employs a Residual Diffusion process that enhances stability and significantly reduces inference time. We performed a critical evaluation against other image-to-image synthesis models, including UNets, GANs, and advanced diffusion models. Our model demonstrates significant improvements in image quality (p<0.05 in MSE, PSNR, and SSIM), inference speed (26 times faster than competing diffusion models), and accurate segmentation results for both nuclei and cell bodies (0.77 and 0.63 mean IOU for nuclei and cell true positives, respectively). This paper is a substantial advancement in the field, providing robust and efficient tools for cell image analysis.
Slot attention aims to decompose an input image into a set of meaningful object files (slots). These latent object representations enable various downstream tasks. Yet, these slots often bind to object parts, not objects themselves, especially for real-world datasets. To address this, we introduce Guided Latent Slot Diffusion - GLASS, an object-centric model that uses generated captions as a guiding signal to better align slots with objects. Our key insight is to learn the slot-attention module in the space of generated images. This allows us to repurpose the pre-trained diffusion decoder model, which reconstructs the images from the slots, as a semantic mask generator based on the generated captions. GLASS learns an object-level representation suitable for multiple tasks simultaneously, e.g., segmentation, image generation, and property prediction, outperforming previous methods. For object discovery, GLASS achieves approx. a +35% and +10% relative improvement for mIoU over the previous state-of-the-art (SOTA) method on the VOC and COCO datasets, respectively, and establishes a new SOTA FID score for conditional image generation amongst slot-attention-based methods. For the segmentation task, GLASS surpasses SOTA weakly-supervised and language-based segmentation models, which were specifically designed for the task.