An Expectation-Maximization Algorithm for Training Clean Diffusion Models from Corrupted Observations
Academy for Advanced Interdisciplinary Studies, Peking University（北京大学前沿交叉学科研究院）；College of Future Technology, Peking University（北京大学未来技术学院）；National Biomedical Imaging Center, Peking University（北京大学国家生物医学成像中心）；Yuanpei College, Peking University（北京大学元培学院）；State Key Laboratory of Multimedia Information Processing, Peking University（北京大学多媒体信息处理国家重点实验室）
Diffusion models excel in solving imaging inverse problems due to their ability to model complex image priors. However, their reliance on large, clean datasets for training limits their practical use where clean data is scarce. In this paper, we propose EMDiffusion, an expectation-maximization (EM) approach to train diffusion models from corrupted observations. Our method alternates between reconstructing clean images from corrupted data using a known diffusion model (E-step) and refining diffusion model weights based on these reconstructions (M-step). This iterative process leads the learned diffusion model to gradually converge to the true clean data distribution. We validate our method through extensive experiments on diverse computational imaging tasks, including random inpainting, denoising, and deblurring, achieving new state-of-the-art performance.
Noise-Robust Conditional Flow Matching: Generating Clean Samples from Noisy Datasets
Center for Advanced Systems Understanding (CASUS)（高级系统理解中心）；Helmholtz-Zentrum Dresden-Rossendorf e. V. (HZDR)（德累斯顿-罗森多夫亥姆霍兹中心）；Institute of Computer Science, University of Wrocław（弗罗茨瓦夫大学计算机科学学院）；School of Computation, Information and Technology, Technical University of Munich（慕尼黑工业大学计算、信息与技术学院）；Cluster of Excellence Physics of Life（生命物理学卓越集群）
Generative adversarial networks (GANs) transform low-dimensional latent vectors into visually plausible images. If the real dataset contains only clean images, then ostensibly, the manifold learned by the GAN should contain only clean images. In this paper, we propose to denoise corrupted images by finding the nearest point on the GAN manifold, recovering latent vectors by minimizing distances in image space. We first demonstrate that given a corrupted version of an image that truly lies on the GAN manifold, we can approximately recover the latent vector and denoise the image, obtaining signif…
The robustness to noise and outliers is an important issue in linear representation in real applications. We focus on the problem that samples are grossly corrupted, which is also the 'sample specific' corruptions problem. A reasonable assumption is that corrupted samples cannot be represented by the dictionary while clean samples can be well represented. This assumption is enforced in this paper by investigating the coefficients of corrupted samples. Concretely, we require the coefficients of corrupted samples be zero. In this way, the representation quality of clean data can be assured witho…
Nankai University（南开大学）；Xi’an Jiaotong University（西安交通大学）；Inception Institute of Artificial Intelligence (IIAI)（人工智能 inception 研究院）；Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)（穆罕默德·本·扎耶德人工智能大学）；Chongqing University（重庆大学）
Supervised deep networks have achieved promisingperformance on image denoising, by learning image priors andnoise statistics on plenty pairs of noisy and clean images. Unsupervised denoising networks are trained with only noisy images. However, for an unseen corrupted image, both supervised andunsupervised networks ignore either its particular image prior, the noise statistics, or both. That is, the networks learned from external images inherently suffer from a domain gap problem: the image priors and noise statistics are very different between the training and test images. This problem become…
Corrupted but Not Broken: Understanding and Mitigating the Negative Impacts of Corrupted Data in Visual Instruction Tuning
Southern University of Science and Technology（南方科技大学）；The Hong Kong University of Science and Technology（香港科技大学）；Huawei Noah’s Ark Lab（华为诺亚实验室）；Hong Kong Baptist University（香港 Baptist 大学）
Deep generative models (e.g. GANs and VAEs) have been developed quite extensively in recent years. Lately, there has been an increased interest in the inversion of such a model, i.e. given a (possibly corrupted) signal, we wish to recover the latent vector that generated it. Building upon sparse representation theory, we define conditions that are applicable to any inversion algorithm (gradient descent, deep encoder, etc.), under which such generative models are invertible with a unique solution. Importantly, the proposed analysis is applicable to any trained model, and does not depend on Gaus…
In this paper, we study color image inpainting as a pure quaternion matrix completion problem. In the literature, the theoretical guarantee for quaternion matrix completion is not well-established. Our main aim is to propose a new minimization problem with an objective combining nuclear norm and a quadratic loss weighted among three channels. To fill the theoretical vacancy, we obtain the error bound in both clean and corrupted regimes, which relies on some new results of quaternion matrices. A general Gaussian noise is considered in robust completion where all observations are corrupted. Moti…
We apply basic statistical reasoning to signal reconstruction by machine learning -- learning to map corrupted observations to clean signals -- with a simple and powerful conclusion: it is possible to learn to restore images by only looking at corrupted examples, at performance at and sometimes exceeding training using clean data, without explicit image priors or likelihood models of the corruption. In practice, we show that a single model learns photographic noise removal, denoising synthetic Monte Carlo images, and reconstruction of undersampled MRI scans -- all corrupted by different proces…
In industrial vision, the anomaly detection problem can be addressed with an autoencoder trained to map an arbitrary image, i.e. with or without any defect, to a clean image, i.e. without any defect. In this approach, anomaly detection relies conventionally on the reconstruction residual or, alternatively, on the reconstruction uncertainty. To improve the sharpness of the reconstruction, we consider an autoencoder architecture with skip connections. In the common scenario where only clean images are available for training, we propose to corrupt them with a synthetic noise model to prevent the …
This paper analyzes various types of image misclassification from a game-theoretic view. Particularly, we consider the misclassification of clean, adversarial, and corrupted images and characterize it through the distribution of multi-order interactions. We discover that the distribution of multi-order interactions varies across the types of misclassification. For example, misclassified adversarial images have a higher strength of high-order interactions than correctly classified clean images, which indicates that adversarial perturbations create spurious features that arise from complex coope…
We treat an image restoration problem with a Poisson noise chan- nel using a Bayesian framework. The Poisson randomness might be appeared in observation of low contrast object in the field of imaging. The noise observation is often hard to treat in a theo- retical analysis. In our formulation, we interpret the observation through the Poisson noise channel as a likelihood, and evaluate the bound of it with a Gaussian function using a latent variable method. We then introduce a Gaussian Markov random field (GMRF) as the prior for the Bayesian approach, and derive the posterior as a Gaussian dist…
University of Electronic Science and Technology of China（电子科技大学）；Tsinghua University（清华大学）；Sichuan University（四川大学）；University of Florida（佛罗里达大学）
Noisy labels composed of correct and corrupted ones are pervasive in practice. They might significantly deteriorate the performance of convolutional neural networks (CNNs), because CNNs are easily overfitted on corrupted labels. To address this issue, inspired by an observation, deep neural networks might first memorize the probably correct-label data and then corrupt-label samples, we propose a novel yet simple self-paced resistance framework to resist corrupted labels, without using any clean validation data. The proposed framework first utilizes the memorization effect of CNNs to learn a cu…
Technion – Israel Institute of Technology（以色列理工学院）
We propose an image processing scheme based on reordering of its patches. For a given corrupted image, we extract all patches with overlaps, refer to these as coordinates in high-dimensional space, and order them such that they are chained in the "shortest possible path", essentially solving the traveling salesman problem. The obtained ordering applied to the corrupted image, implies a permutation of the image pixels to what should be a regular signal. This enables us to obtain good recovery of the clean image by applying relatively simple 1D smoothing operations (such as filtering or interpol…
Recent studies on learning-based image denoising have achieved promising performance on various noise reduction tasks. Most of these deep denoisers are trained either under the supervision of clean references, or unsupervised on synthetic noise. The assumption with the synthetic noise leads to poor generalization when facing real photographs. To address this issue, we propose a novel deep image-denoising method by regarding the noise reduction task as a special case of the noise transference task. Learning noise transference enables the network to acquire the denoising ability by observing the…
Deep convolutional neural networks have driven substantial advancements in the automatic understanding of images. Requiring a large collection of images and their associated annotations is one of the main bottlenecks limiting the adoption of deep networks. In the task of medical image segmentation, requiring pixel-level semantic annotations performed by human experts exacerbate this difficulty. This paper proposes a new framework to train a fully convolutional segmentation network from a large set of cheap unreliable annotations and a small set of expert-level clean annotations. We propose a s…
Insight SFI Centre for Data Analytics（Insight SFI数据分析中心）；Dublin City University（都柏林城市大学）
Using search engines for web image retrieval is a tempting alternative to manual curation when creating an image dataset, but their main drawback remains the proportion of incorrect (noisy) samples retrieved. These noisy samples have been evidenced by previous works to be a mixture of in-distribution (ID) samples, assigned to the incorrect category but presenting similar visual semantics to other classes in the dataset, and out-of-distribution (OOD) images, which share no semantic correlation with any category from the dataset. The latter are, in practice, the dominant type of noisy images ret…
ByteDance Inc（字节跳动公司）；Peking University Shenzhen Graduate School（北京大学深圳研究生院）
Semi-supervised learning based methods are current SOTA solutions to the noisy-label learning problem, which rely on learning an unsupervised label cleaner first to divide the training samples into a labeled set for clean data and an unlabeled set for noise data. Typically, the cleaner is obtained via fitting a mixture model to the distribution of per-sample training losses. However, the modeling procedure is \emph{class agnostic} and assumes the loss distributions of clean and noise samples are the same across different classes. Unfortunately, in practice, such an assumption does not always h…
When designing a diagnostic model for a clinical application, it is crucial to guarantee the robustness of the model with respect to a wide range of image corruptions. Herein, an easy-to-use benchmark is established to evaluate how deep neural networks perform on corrupted pathology images. Specifically, corrupted images are generated by injecting nine types of common corruptions into validation images. Besides, two classification and one ranking metrics are designed to evaluate the prediction and confidence performance under corruption. Evaluated on two resulting benchmark datasets, we find t…
Yau Mathematical Sciences Center, Tsinghua University（清华大学丘成桐数学科学中心）；Hisilicon（海思）；Institute for Interdisciplinary Information Science, Tsinghua University（清华大学交叉信息研究院）；Yanqi Lake Beijing Institute of Mathematical Sciences and Applications（北京雁栖湖应用数学研究院）
Collecting paired training data is difficult in practice, but the unpaired samples broadly exist. Current approaches aim at generating synthesized training data from unpaired samples by exploring the relationship between the corrupted and clean data. This work proposes LUD-VAE, a deep generative method to learn the joint probability density function from data sampled from marginal distributions. Our approach is based on a carefully designed probabilistic graphical model in which the clean and corrupted data domains are conditionally independent. Using variational inference, we maximize the evi…
We study a new family of inverse problems for recovering representations of corrupted data. We assume access to a pre-trained representation learning network R(x) that operates on clean images, like CLIP. The problem is to recover the representation of an image R(x), if we are only given a corrupted version A(x), for some known forward operator A. We propose a supervised inversion method that uses a contrastive objective to obtain excellent representations for highly corrupted images. Using a linear probe on our robust representations, we achieve a higher accuracy than end-to-end supervised ba…
University of California, Berkeley（加州大学伯克利分校）；University of Chicago（芝加哥大学）；Foundational Research Institute（基础研究所）；Toyota Technological Institute at Chicago（芝加哥丰田理工学院）
The growing importance of massive datasets used for deep learning makes robustness to label noise a critical property for classifiers to have. Sources of label noise include automatic labeling, non-expert labeling, and label corruption by data poisoning adversaries. Numerous previous works assume that no source of labels can be trusted. We relax this assumption and assume that a small subset of the training data is trusted. This enables substantial label corruption robustness performance gains. In addition, particularly severe label noise can be combated by using a set of trusted data with cle…
The real-world facial expression recognition (FER) datasets suffer from noisy annotations due to crowd-sourcing, ambiguity in expressions, the subjectivity of annotators and inter-class similarity. However, the recent deep networks have strong capacity to memorize the noisy annotations leading to corrupted feature embedding and poor generalization. To handle noisy annotations, we propose a dynamic FER learning framework (DNFER) in which clean samples are selected based on dynamic class specific threshold during training. Specifically, DNFER is based on supervised training using selected clean …
School of Electrical and Computer Engineering, Purdue University（普渡大学电气与计算机工程学院）
Feature-based student-teacher learning, a training method that encourages the student's hidden features to mimic those of the teacher network, is empirically successful in transferring the knowledge from a pre-trained teacher network to the student network. Furthermore, recent empirical results demonstrate that, the teacher's features can boost the student network's generalization even when the student's input sample is corrupted by noise. However, there is a lack of theoretical insights into why and when this method of transferring knowledge can be successful between such heterogeneous tasks.…
IIIT-Delhi（德里印度理工学院）；Arizona State University（亚利桑那州立大学）
Input transformation based defense strategies fall short in defending against strong adversarial attacks. Some successful defenses adopt approaches that either increase the randomness within the applied transformations, or make the defense computationally intensive, making it substantially more challenging for the attacker. However, it limits the applicability of such defenses as a pre-processing step, similar to computationally heavy approaches that use retraining and network modifications to achieve robustness to perturbations. In this work, we propose a defense strategy that applies random …
Korea Advanced Institute of Science and Technology (KAIST)（韩国科学技术院）；Kim Jaechul Graduate School of AI（金在哲人工智能研究生院）
Recently, there has been extensive research interest in training deep networks to denoise images without clean reference. However, the representative approaches such as Noise2Noise, Noise2Void, Stein's unbiased risk estimator (SURE), etc. seem to differ from one another and it is difficult to find the coherent mathematical structure. To address this, here we present a novel approach, called Noise2Score, which reveals a missing link in order to unite these seemingly different approaches. Specifically, we show that image denoising problems without clean images can be addressed by finding the mod…
We describe a novel method for training high-quality image denoising models based on unorganized collections of corrupted images. The training does not need access to clean reference images, or explicit pairs of corrupted images, and can thus be applied in situations where such data is unacceptably expensive or impossible to acquire. We build on a recent technique that removes the need for reference data by employing networks with a "blind spot" in the receptive field, and significantly improve two key aspects: image quality and training efficiency. Our result quality is on par with state-of-t…
Technical University Munich（慕尼黑工业大学）；OneProjects Design Innovation GmbH（OneProjects设计创新有限责任公司）
In ultrasound imaging the appearance of homogeneous regions of tissue is subject to speckle, which for certain applications can make the detection of tissue irregularities difficult. To cope with this, it is common practice to apply speckle reduction filters to the images. Most conventional filtering techniques are fairly hand-crafted and often need to be finely tuned to the present hardware, imaging scheme and application. Learning based techniques on the other hand suffer from the need for a target image for training (in case of fully supervised techniques) or require narrow, complex physics…
