Search

Multimodal learning seeks to combine data from multiple input sources to enhance the performance of different downstream tasks. In real-world scenarios, performance can degrade substantially if some input modalities are missing. Existing methods that can handle missing modalities involve custom training or adaptation steps for each input modality combination. These approaches are either tied to specific modalities or become computationally expensive as the number of input modalities increases. In this paper, we propose Masked Modality Projection (MMP), a method designed to train a single model that is robust to any missing modality scenario. We achieve this by randomly masking a subset of modalities during training and learning to project available input modalities to estimate the tokens for the masked modalities. This approach enables the model to effectively learn to leverage the information from the available modalities to compensate for the missing ones, enhancing missing modality robustness. We conduct a series of experiments with various baseline models and datasets to assess the effectiveness of this strategy. Experiments demonstrate that our approach improves robustness to different missing modality scenarios, outperforming existing methods designed for missing modalities or specific modality combinations.

Diffusion models can generate a variety of high-quality images by modeling complex data distributions. Trained diffusion models can also be very effective image priors for solving inverse problems. Most of the existing diffusion-based methods integrate data consistency steps within the diffusion reverse sampling process. The data consistency steps rely on an approximate likelihood function. In this paper, we show that the existing approximations are either insufficient or computationally inefficient. To address these issues, we propose a unified likelihood approximation method that incorporates a covariance correction term to enhance the performance and avoids propagating gradients through the diffusion model. The correction term, when integrated into the reverse diffusion sampling process, achieves better convergence towards the true data posterior for selected distributions and improves performance on real-world natural image datasets. Furthermore, we present an efficient way to factorize and invert the covariance matrix of the likelihood function for several inverse problems. We present comprehensive experiments to demonstrate the effectiveness of our method over several existing approaches.

Unauthorized privacy-related and copyrighted content generation using generative-AI is becoming a significant concern for human society, raising ethical, legal, and privacy issues that demand urgent attention. The EU's General Data Protection Regulation (GDPR) include a "right to be forgotten," which allows individuals to request the deletion of their personal data. However, this primarily applies to data stored in traditional databases, not AI models. Recently, machine unlearning techniques have arise that attempt to eliminate the influence of sensitive content used during AI model training, but they often require extensive updates to the deployed systems and incur substantial computational costs. In this work, we propose a novel and efficient method called Single Layer Unlearning Gradient (SLUG), that can unlearn targeted information by updating targeted layers of a model using a one-time gradient computation. Our method is highly modular and enables the selective removal of multiple sensitive concepts, such as celebrity names and copyrighted content, from the generated outputs of widely used foundation models (e.g., CLIP) and generative models (e.g., Stable Diffusion). Broadly, our method ensures AI-generated content complies with privacy regulations and intellectual property laws, fostering responsible use of generative models, mitigating legal risks and promoting a trustworthy, socially responsible AI ecosystem.

We introduce transformation-dependent adversarial attacks, a new class of threats where a single additive perturbation can trigger diverse, controllable mis-predictions by systematically transforming the input (e.g., scaling, blurring, compression). Unlike traditional attacks with static effects, our perturbations embed metamorphic properties to enable different adversarial attacks as a function of the transformation parameters. We demonstrate the transformation-dependent vulnerability across models (e.g., convolutional networks and vision transformers) and vision tasks (e.g., image classification and object detection). Our proposed geometric and photometric transformations enable a range of targeted errors from one crafted input (e.g., higher than 90% attack success rate for classifiers). We analyze effects of model architecture and type/variety of transformations on attack effectiveness. This work forces a paradigm shift by redefining adversarial inputs as dynamic, controllable threats. We highlight the need for robust defenses against such multifaceted, chameleon-like perturbations that current techniques are ill-prepared for.

Recent studies reveal that integrating new modalities into Large Language Models (LLMs), such as Vision-Language Models (VLMs), creates a new attack surface that bypasses existing safety training techniques like Supervised Fine-tuning (SFT) and Reinforcement Learning with Human Feedback (RLHF). While further SFT and RLHF-based safety training can be conducted in multi-modal settings, collecting multi-modal training datasets poses a significant challenge. Inspired by the structural design of recent multi-modal models, where, regardless of the combination of input modalities, all inputs are ultimately fused into the language space, we aim to explore whether unlearning solely in the textual domain can be effective for cross-modality safety alignment. Our evaluation across six datasets empirically demonstrates the transferability -- textual unlearning in VLMs significantly reduces the Attack Success Rate (ASR) to less than 8\% and in some cases, even as low as nearly 2\% for both text-based and vision-text-based attacks, alongside preserving the utility. Moreover, our experiments show that unlearning with a multi-modal dataset offers no potential benefits but incurs significantly increased computational demands, possibly up to 6 times higher.

The primary focus of Neural Representation for Videos (NeRV) is to effectively model its spatiotemporal consistency. However, current NeRV systems often face a significant issue of spatial inconsistency, leading to decreased perceptual quality. To address this issue, we introduce the Pyramidal Neural Representation for Videos (PNeRV), which is built on a multi-scale information connection and comprises a lightweight rescaling operator, Kronecker Fully-connected layer (KFc), and a Benign Selective Memory (BSM) mechanism. The KFc, inspired by the tensor decomposition of the vanilla Fully-connected layer, facilitates low-cost rescaling and global correlation modeling. BSM merges high-level features with granular ones adaptively. Furthermore, we provide an analysis based on the Universal Approximation Theory of the NeRV system and validate the effectiveness of the proposed PNeRV.We conducted comprehensive experiments to demonstrate that PNeRV surpasses the performance of contemporary NeRV models, achieving the best results in video regression on UVG and DAVIS under various metrics (PSNR, SSIM, LPIPS, and FVD). Compared to vanilla NeRV, PNeRV achieves a +4.49 dB gain in PSNR and a 231% increase in FVD on UVG, along with a +3.28 dB PSNR and 634% FVD increase on DAVIS.

Plug-and-Play Priors (PnP) is a well-known class of methods for solving inverse problems in computational imaging. PnP methods combine physical forward models with learned prior models specified as image denoisers. A common issue with the learned models is that of a performance drop when there is a distribution shift between the training and testing data. Test-time training (TTT) was recently proposed as a general strategy for improving the performance of learned models when training and testing data come from different distributions. In this paper, we propose PnP-TTT as a new method for overcoming distribution shifts in PnP. PnP-TTT uses deep equilibrium learning (DEQ) for optimizing a self-supervised loss at the fixed points of PnP iterations. PnP-TTT can be directly applied on a single test sample to improve the generalization of PnP. We show through simulations that given a sufficient number of measurements, PnP-TTT enables the use of image priors trained on natural images for image reconstruction in magnetic resonance imaging (MRI).

The capability to accurately estimate 3D human poses is crucial for diverse fields such as action recognition, gait recognition, and virtual/augmented reality. However, a persistent and significant challenge within this field is the accurate prediction of human poses under conditions of severe occlusion. Traditional image-based estimators struggle with heavy occlusions due to a lack of temporal context, resulting in inconsistent predictions. While video-based models benefit from processing temporal data, they encounter limitations when faced with prolonged occlusions that extend over multiple frames. This challenge arises because these models struggle to generalize beyond their training datasets, and the variety of occlusions is hard to capture in the training data. Addressing these challenges, we propose STRIDE (Single-video based TempoRally contInuous Occlusion-Robust 3D Pose Estimation), a novel Test-Time Training (TTT) approach to fit a human motion prior for each video. This approach specifically handles occlusions that were not encountered during the model's training. By employing STRIDE, we can refine a sequence of noisy initial pose estimates into accurate, temporally coherent poses during test time, effectively overcoming the limitations of prior methods. Our framework demonstrates flexibility by being model-agnostic, allowing us to use any off-the-shelf 3D pose estimation method for improving robustness and temporal consistency. We validate STRIDE's efficacy through comprehensive experiments on challenging datasets like Occluded Human3.6M, Human3.6M, and OCMotion, where it not only outperforms existing single-image and video-based pose estimation models but also showcases superior handling of substantial occlusions, achieving fast, robust, accurate, and temporally consistent 3D pose estimates. Code is made publicly available at https://github.com/take2rohit/stride, Comment: Paper accepted at IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)-2025

Deep learning-based methods deliver state-of-the-art performance for solving inverse problems that arise in computational imaging. These methods can be broadly divided into two groups: (1) learn a network to map measurements to the signal estimate, which is known to be fragile; (2) learn a prior for the signal to use in an optimization-based recovery. Despite the impressive results from the latter approach, many of these methods also lack robustness to shifts in data distribution, measurements, and noise levels. Such domain shifts result in a performance gap and in some cases introduce undesired artifacts in the estimated signal. In this paper, we explore the qualitative and quantitative effects of various domain shifts and propose a flexible and parameter efficient framework that adapt pretrained networks to such shifts. We demonstrate the effectiveness of our method for a number of natural image, MRI, and CT reconstructions tasks under domain, measurement model, and noise-level shifts. Our experiments demonstrate that our method provides significantly better performance and parameter efficiency compared to existing domain adaptation techniques.

Multi-task and multi-domain learning methods seek to learn multiple tasks/domains, jointly or one after another, using a single unified network. The key challenge and opportunity is to exploit shared information across tasks and domains to improve the efficiency of the unified network. The efficiency can be in terms of accuracy, storage cost, computation, or sample complexity. In this paper, we propose a factorized tensor network (FTN) that can achieve accuracy comparable to independent single-task/domain networks with a small number of additional parameters. FTN uses a frozen backbone network from a source model and incrementally adds task/domain-specific low-rank tensor factors to the shared frozen network. This approach can adapt to a large number of target domains and tasks without catastrophic forgetting. Furthermore, FTN requires a significantly smaller number of task-specific parameters compared to existing methods. We performed experiments on widely used multi-domain and multi-task datasets. We show the experiments on convolutional-based architecture with different backbones and on transformer-based architecture. We observed that FTN achieves similar accuracy as single-task/domain methods while using only a fraction of additional parameters per task.

Multimodal learning seeks to utilize data from multiple sources to improve the overall performance of downstream tasks. It is desirable for redundancies in the data to make multimodal systems robust to missing or corrupted observations in some correlated modalities. However, we observe that the performance of several existing multimodal networks significantly deteriorates if one or multiple modalities are absent at test time. To enable robustness to missing modalities, we propose a simple and parameter-efficient adaptation procedure for pretrained multimodal networks. In particular, we exploit modulation of intermediate features to compensate for the missing modalities. We demonstrate that such adaptation can partially bridge performance drop due to missing modalities and outperform independent, dedicated networks trained for the available modality combinations in some cases. The proposed adaptation requires extremely small number of parameters (e.g., fewer than 1% of the total parameters) and applicable to a wide range of modality combinations and tasks. We conduct a series of experiments to highlight the missing modality robustness of our proposed method on five different multimodal tasks across seven datasets. Our proposed method demonstrates versatility across various tasks and datasets, and outperforms existing methods for robust multimodal learning with missing modalities., Comment: Accepted by IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). 28 pages, 6 figures, 17 tables

Plug-and-Play (PnP) priors is a widely-used family of methods for solving imaging inverse problems by integrating physical measurement models with image priors specified using image denoisers. PnP methods have been shown to achieve state-of-the-art performance when the prior is obtained using powerful deep denoisers. Despite extensive work on PnP, the topic of distribution mismatch between the training and testing data has often been overlooked in the PnP literature. This paper presents a set of new theoretical and numerical results on the topic of prior distribution mismatch and domain adaptation for alternating direction method of multipliers (ADMM) variant of PnP. Our theoretical result provides an explicit error bound for PnP-ADMM due to the mismatch between the desired denoiser and the one used for inference. Our analysis contributes to the work in the area by considering the mismatch under nonconvex data-fidelity terms and expansive denoisers. Our first set of numerical results quantifies the impact of the prior distribution mismatch on the performance of PnP-ADMM on the problem of image super-resolution. Our second set of numerical results considers a simple and effective domain adaption strategy that closes the performance gap due to the use of mismatched denoisers. Our results suggest the relative robustness of PnP-ADMM to prior distribution mismatch, while also showing that the performance gap can be significantly reduced with few training samples from the desired distribution.

Leveraging information across diverse modalities is known to enhance performance on multimodal segmentation tasks. However, effectively fusing information from different modalities remains challenging due to the unique characteristics of each modality. In this paper, we propose a novel fusion strategy that can effectively fuse information from different modality combinations. We also propose a new model named Multi-Modal Segmentation TransFormer (MMSFormer) that incorporates the proposed fusion strategy to perform multimodal material and semantic segmentation tasks. MMSFormer outperforms current state-of-the-art models on three different datasets. As we begin with only one input modality, performance improves progressively as additional modalities are incorporated, showcasing the effectiveness of the fusion block in combining useful information from diverse input modalities. Ablation studies show that different modules in the fusion block are crucial for overall model performance. Furthermore, our ablation studies also highlight the capacity of different input modalities to improve performance in the identification of different types of materials. The code and pretrained models will be made available at https://github.com/csiplab/MMSFormer., Comment: Accepted by IEEE Open Journal of Signal Processing. 15 pages, 3 figures, 9 tables

Existing implicit neural representation (INR) methods do not fully exploit spatiotemporal redundancies in videos. Index-based INRs ignore the content-specific spatial features and hybrid INRs ignore the contextual dependency on adjacent frames, leading to poor modeling capability for scenes with large motion or dynamics. We analyze this limitation from the perspective of function fitting and reveal the importance of frame difference. To use explicit motion information, we propose Difference Neural Representation for Videos (DNeRV), which consists of two streams for content and frame difference. We also introduce a collaborative content unit for effective feature fusion. We test DNeRV for video compression, inpainting, and interpolation. DNeRV achieves competitive results against the state-of-the-art neural compression approaches and outperforms existing implicit methods on downstream inpainting and interpolation for $960 \times 1920$ videos.

Digital radiography (DR) is becoming popular for the point of care imaging in the recent past. To reduce the radiation exposure, controlled radiation based on as low as reasonably achievable (ALARA) principle is employed and this results in low contrast images. To address this issue, post-processing algorithms such as the Multiscale Image Contrast Amplification (MUSICA) algorithm can be used to enhance the contrast of DR images even with a low radiation dose. In this study, a modification of the MUSICA algorithm is investigated to determine the potential for further contrast improvement specifically for DR images. The conclusion is that combining log compression and its inverse at the appropriate stage with a multi-stage MUSICA and denoising is very promising. The proposed method resulted in an average of 66.5 % increase in the mean contrast-to-noise ratio (CNR) for the test images considered., Comment: 4 pages, 4 figures

Conventional ultrasound (US) imaging employs the delay and sum (DAS) receive beamforming with dynamic receive focus for image reconstruction due to its simplicity and robustness. However, the DAS beamforming follows a geometrical method of delay estimation with a spatially constant speed-of-sound (SoS) of 1540 m/s throughout the medium irrespective of the tissue in-homogeneity. This approximation leads to errors in delay estimations that accumulate with depth and degrades the resolution, contrast and overall accuracy of the US image. In this work, we propose a fast marching based DAS for focused transmissions which leverages the approximate SoS map to estimate the refraction corrected propagation delays for each pixel in the medium. The proposed approach is validated qualitatively and quantitatively for imaging depths of upto ~ 11 cm through simulations, where fat layer induced aberration is employed to alter the SoS in the medium. To the best of authors' knowledge, this is the first work considering the effect of SoS on image quality for deeper imaging., Comment: 5 pages, 4 figures

We propose an approach for adversarial attacks on dense prediction models (such as object detectors and segmentation). It is well known that the attacks generated by a single surrogate model do not transfer to arbitrary (blackbox) victim models. Furthermore, targeted attacks are often more challenging than the untargeted attacks. In this paper, we show that a carefully designed ensemble can create effective attacks for a number of victim models. In particular, we show that normalization of the weights for individual models plays a critical role in the success of the attacks. We then demonstrate that by adjusting the weights of the ensemble according to the victim model can further improve the performance of the attacks. We performed a number of experiments for object detectors and segmentation to highlight the significance of the our proposed methods. Our proposed ensemble-based method outperforms existing blackbox attack methods for object detection and segmentation. Finally we show that our proposed method can also generate a single perturbation that can fool multiple blackbox detection and segmentation models simultaneously. Code is available at https://github.com/CSIPlab/EBAD., Comment: CVPR 2023 Accepted

With the rise of cameras and smart sensors, humanity generates an exponential amount of data. This valuable information, including underrepresented cases like AI in medical settings, can fuel new deep-learning tools. However, data scientists must prioritize ensuring privacy for individuals in these untapped datasets, especially for images or videos with faces, which are prime targets for identification methods. Proposed solutions to de-identify such images often compromise non-identifying facial attributes relevant to downstream tasks. In this paper, we introduce Disguise, a novel algorithm that seamlessly de-identifies facial images while ensuring the usability of the modified data. Unlike previous approaches, our solution is firmly grounded in the domains of differential privacy and ensemble-learning research. Our method involves extracting and substituting depicted identities with synthetic ones, generated using variational mechanisms to maximize obfuscation and non-invertibility. Additionally, we leverage supervision from a mixture-of-experts to disentangle and preserve other utility attributes. We extensively evaluate our method using multiple datasets, demonstrating a higher de-identification rate and superior consistency compared to prior approaches in various downstream tasks., Comment: Accepted at AAAI 2024. Paper + supplementary material

Deep networks can be trained to map images into a low-dimensional latent space. In many cases, different images in a collection are articulated versions of one another; for example, same object with different lighting, background, or pose. Furthermore, in many cases, parts of images can be corrupted by noise or missing entries. In this paper, our goal is to recover images without access to the ground-truth (clean) images using the articulations as structural prior of the data. Such recovery problems fall under the domain of compressive sensing. We propose to learn autoencoder with tensor ring factorization on the the embedding space to impose structural constraints on the data. In particular, we use a tensor ring structure in the bottleneck layer of the autoencoder that utilizes the soft labels of the structured dataset. We empirically demonstrate the effectiveness of the proposed approach for inpainting and denoising applications. The resulting method achieves better reconstruction quality compared to other generative prior-based self-supervised recovery approaches for compressive sensing.

Mask-based lensless cameras offer a novel design for imaging systems by replacing the lens in a conventional camera with a layer of coded mask. Each pixel of the lensless camera encodes the information of the entire 3D scene. Existing methods for 3D reconstruction from lensless measurements suffer from poor spatial and depth resolution. This is partially due to the system ill conditioning that arises because the point-spread functions (PSFs) from different depth planes are very similar. In this paper, we propose to capture multiple measurements of the scene under a sequence of coded illumination patterns to improve the 3D image reconstruction quality. In addition, we put the illumination source at a distance away from the camera. With such baseline distance between the lensless camera and illumination source, the camera observes a slice of the 3D volume, and the PSF of each depth plane becomes more resolvable from each other. We present simulation results along with experimental results with a camera prototype to demonstrate the effectiveness of our approach.