Search

We show that deep neural networks achieve dimension-independent rates of convergence for learning structured densities such as those arising in image, audio, video, and text applications. More precisely, we demonstrate that neural networks with a simple $L^2$-minimizing loss achieve a rate of $n^{-1/(4+r)}$ in nonparametric density estimation when the underlying density is Markov to a graph whose maximum clique size is at most $r$, and we provide evidence that in the aforementioned applications, this size is typically constant, i.e., $r=O(1)$. We then establish that the optimal rate in $L^1$ is $n^{-1/(2+r)}$ which, compared to the standard nonparametric rate of $n^{-1/(2+d)}$, reveals that the effective dimension of such problems is the size of the largest clique in the Markov random field. These rates are independent of the data's ambient dimension, making them applicable to realistic models of image, sound, video, and text data. Our results provide a novel justification for deep learning's ability to circumvent the curse of dimensionality, demonstrating dimension-independent convergence rates in these contexts.

In semi-supervised domain adaptation (SSDA), the model aims to leverage partially labeled target domain data along with a large amount of labeled source domain data to enhance its generalization capability for the target domain. A key advantage of SSDA is its ability to significantly reduce reliance on labeled data, thereby lowering the costs and time associated with data preparation. Most existing SSDA methods utilize information from domain labels and class labels but overlook the structural information of the data. To address this issue, this paper proposes a graph learning perspective (AGLP) for semi-supervised domain adaptation. We apply the graph convolutional network to the instance graph which allows structural information to propagate along the weighted graph edges. The proposed AGLP model has several advantages. First, to the best of our knowledge, this is the first work to model structural information in SSDA. Second, the proposed model can effectively learn domain-invariant and semantic representations, reducing domain discrepancies in SSDA. Extensive experimental results on multiple standard benchmarks demonstrate that the proposed AGLP algorithm outperforms state-of-the-art semi-supervised domain adaptation methods., Comment: 8page

Semi-supervised learning (SSL) has made notable advancements in medical image segmentation (MIS), particularly in scenarios with limited labeled data and significantly enhancing data utilization efficiency. Previous methods primarily focus on complex training strategies to utilize unlabeled data but neglect the importance of graph structural information. Different from existing methods, we propose a graph-based clustering for semi-supervised medical image segmentation (GraphCL) by jointly modeling graph data structure in a unified deep model. The proposed GraphCL model enjoys several advantages. Firstly, to the best of our knowledge, this is the first work to model the data structure information for semi-supervised medical image segmentation (SSMIS). Secondly, to get the clustered features across different graphs, we integrate both pairwise affinities between local image features and raw features as inputs. Extensive experimental results on three standard benchmarks show that the proposed GraphCL algorithm outperforms state-of-the-art semi-supervised medical image segmentation methods., Comment: 9page

As a technique to alleviate the pressure of data annotation, semi-supervised learning (SSL) has attracted widespread attention. In the specific domain of medical image segmentation, semi-supervised methods (SSMIS) have become a research hotspot due to their ability to reduce the need for large amounts of precisely annotated data. SSMIS focuses on enhancing the model's generalization performance by leveraging a small number of labeled samples and a large number of unlabeled samples. The latest sharpness-aware optimization (SAM) technique, which optimizes the model by reducing the sharpness of the loss function, has shown significant success in SSMIS. However, SAM and its variants may not fully account for the distribution differences between different datasets. To address this issue, we propose a sharpness-aware optimization method based on $f$-divergence minimization (DiM) for semi-supervised medical image segmentation. This method enhances the model's stability by fine-tuning the sensitivity of model parameters and improves the model's adaptability to different datasets through the introduction of $f$-divergence. By reducing $f$-divergence, the DiM method not only improves the performance balance between the source and target datasets but also prevents performance degradation due to overfitting on the source dataset., Comment: 8page

Our study introduces ResNet-L2 (RL2), a novel metric for evaluating generative models and image quality in histopathology, addressing limitations of traditional metrics, such as Frechet inception distance (FID), when the data is scarce. RL2 leverages ResNet features with a normalizing flow to calculate RMSE distance in the latent space, providing reliable assessments across diverse histopathology datasets. We evaluated the performance of RL2 on degradation types, such as blur, Gaussian noise, salt-and-pepper noise, and rectangular patches, as well as diffusion processes. RL2's monotonic response to increasing degradation makes it well-suited for models that assess image quality, proving a valuable advancement for evaluating image generation techniques in histopathology. It can also be used to discard low-quality patches while sampling from a whole slide image. It is also significantly lighter and faster compared to traditional metrics and requires fewer images to give stable metric value., Comment: 7 pages, 5 figures

Deep neural networks can obtain impressive performance on various tasks under the assumption that their training domain is identical to their target domain. Performance can drop dramatically when this assumption does not hold. One explanation for this discrepancy is the presence of spurious domain-specific correlations in the training data that the network exploits. Causal mechanisms, in the other hand, can be made invariant under distribution changes as they allow disentangling the factors of distribution underlying the data generation. Yet, learning causal mechanisms to improve out-of-distribution generalisation remains an under-explored area. We propose a Bayesian neural architecture that disentangles the learning of the the data distribution from the inference process mechanisms. We show theoretically and experimentally that our model approximates reasoning under causal interventions. We demonstrate the performance of our method, outperforming point estimate-counterparts, on out-of-distribution image recognition tasks where the data distribution acts as strong adversarial confounders., Comment: 16 pages, 10 pages for main paper and 6 pages for references and appendix, 8 figures

Point spread function (PSF) engineering is vital for precisely controlling the focus of light in computational imaging, with applications in neural imaging, fluorescence microscopy, and biophotonics. The PSF is derived from the magnitude of the Fourier transform of a phase function, making the construction of the phase function given the PSF (PSF engineering) an ill-posed inverse problem. Traditional PSF engineering methods rely on physical basis functions, limiting their ability to generalize across the range of PSFs required for imaging tasks. We introduce a novel approach leveraging implicit neural representations that significantly outperforms existing pixel-wise optimization methods in phase function quality., Comment: 3 pages, 7 figures. To be published in ICVISP 2024 (https://www.icvisp.org/)

Long-tailed problems in healthcare emerge from data imbalance due to variability in the prevalence and representation of different medical conditions, warranting the requirement of precise and dependable classification methods. Traditional loss functions such as cross-entropy and binary cross-entropy are often inadequate due to their inability to address the imbalances between the classes with high representation and the classes with low representation found in medical image datasets. We introduce a novel polynomial loss function based on Pade approximation, designed specifically to overcome the challenges associated with long-tailed classification. This approach incorporates asymmetric sampling techniques to better classify under-represented classes. We conducted extensive evaluations on three publicly available medical datasets and a proprietary medical dataset. Our implementation of the proposed loss function is open-sourced in the public repository:https://github.com/ipankhi/ALPA., Comment: 13 pages, 1 figures. Accepted in The 35th British Machine Vision Conference (BMVC24)

Federated Learning (FL) facilitates data privacy by enabling collaborative in-situ training across decentralized clients. Despite its inherent advantages, FL faces significant challenges of performance and convergence when dealing with data that is not independently and identically distributed (non-i.i.d.). While previous research has primarily addressed the issue of skewed label distribution across clients, this study focuses on the less explored challenge of multi-domain FL, where client data originates from distinct domains with varying feature distributions. We introduce a novel method designed to address these challenges FedStein: Enhancing Multi-Domain Federated Learning Through the James-Stein Estimator. FedStein uniquely shares only the James-Stein (JS) estimates of batch normalization (BN) statistics across clients, while maintaining local BN parameters. The non-BN layer parameters are exchanged via standard FL techniques. Extensive experiments conducted across three datasets and multiple models demonstrate that FedStein surpasses existing methods such as FedAvg and FedBN, with accuracy improvements exceeding 14% in certain domains leading to enhanced domain generalization. The code is available at https://github.com/sunnyinAI/FedStein, Comment: 12 pages, 2 figures. Accepted at International Workshop on Federated Foundation Models In Conjunction with NeurIPS 2024 (FL@FM-NeurIPS'24)

We propose two new evaluation metrics to assess realness of generated images based on normalizing flows: a simpler and efficient flow-based likelihood distance (FLD) and a more exact dual-flow based likelihood distance (D-FLD). Because normalizing flows can be used to compute the exact likelihood, the proposed metrics assess how closely generated images align with the distribution of real images from a given domain. This property gives the proposed metrics a few advantages over the widely used Fr\'echet inception distance (FID) and other recent metrics. Firstly, the proposed metrics need only a few hundred images to stabilize (converge in mean), as opposed to tens of thousands needed for FID, and at least a few thousand for the other metrics. This allows confident evaluation of even small sets of generated images, such as validation batches inside training loops. Secondly, the network used to compute the proposed metric has over an order of magnitude fewer parameters compared to Inception-V3 used to compute FID, making it computationally more efficient. For assessing the realness of generated images in new domains (e.g., x-ray images), ideally these networks should be retrained on real images to model their distinct distributions. Thus, our smaller network will be even more advantageous for new domains. Extensive experiments show that the proposed metrics have the desired monotonic relationships with the extent of image degradation of various kinds., Comment: 15 pages, 16 figures

Autism spectrum disorders (ASDs) are developmental conditions characterized by restricted interests and difficulties in communication. The complexity of ASD has resulted in a deficiency of objective diagnostic biomarkers. Deep learning methods have gained recognition for addressing these challenges in neuroimaging analysis, but finding and interpreting such diagnostic biomarkers are still challenging computationally. We propose an ASD feature reduction pipeline using resting-state fMRI (rs-fMRI). We used Ncuts parcellations and Power atlas to extract functional connectivity data, resulting in over 30 thousand features. Then the pipeline further compresses the connectivities into 5 latent Gaussian distributions, providing is a low-dimensional representation of the data, using a denoising variational autoencoder (DVAE). To test the method, we employed the extracted latent features from the DVAE to classify ASD using traditional classifiers such as support vector machine (SVM) on a large multi-site dataset. The 95% confidence interval for the prediction accuracy of the SVM is [0.63, 0.76] after site harmonization using the extracted latent distributions. Without using DVAE, the prediction accuracy is 0.70, which falls within the interval. This implies that the model successfully encodes the diagnostic information in rs-fMRI data to 5 Gaussian distributions (10 features) without sacrificing prediction performance. The runtime for training the DVAE and obtaining classification results from its extracted latent features (37 minutes) was 7 times shorter compared to training classifiers directly on the raw connectivity matrices (5-6 hours). Our findings also suggest that the Power atlas provides more effective brain connectivity insights for diagnosing ASD than Ncuts parcellations. The encoded features can be used for the help of diagnosis and interpretation of the disease.

Learning a discriminative model to distinguish a target from its surrounding distractors is essential to generic visual object tracking. Dynamic target representation adaptation against distractors is challenging due to the limited discriminative capabilities of prevailing trackers. We present a new visual Prompting mechanism for generic Visual Object Tracking (PiVOT) to address this issue. PiVOT proposes a prompt generation network with the pre-trained foundation model CLIP to automatically generate and refine visual prompts, enabling the transfer of foundation model knowledge for tracking. While CLIP offers broad category-level knowledge, the tracker, trained on instance-specific data, excels at recognizing unique object instances. Thus, PiVOT first compiles a visual prompt highlighting potential target locations. To transfer the knowledge of CLIP to the tracker, PiVOT leverages CLIP to refine the visual prompt based on the similarities between candidate objects and the reference templates across potential targets. Once the visual prompt is refined, it can better highlight potential target locations, thereby reducing irrelevant prompt information. With the proposed prompting mechanism, the tracker can generate improved instance-aware feature maps through the guidance of the visual prompt, thus effectively reducing distractors. The proposed method does not involve CLIP during training, thereby keeping the same training complexity and preserving the generalization capability of the pretrained foundation model. Extensive experiments across multiple benchmarks indicate that PiVOT, using the proposed prompting method can suppress distracting objects and enhance the tracker., Comment: Accepted and to appear in IEEE Transactions on Multimedia

Inconsistent responses of X-ray detector elements lead to stripe artifacts in the sinogram data, which manifest as ring artifacts in the reconstructed CT images, severely degrading image quality. This paper proposes a method for correcting stripe artifacts in the sinogram data. The proposed method leverages implicit neural representation (INR) to correct defective pixel response values using implicit continuous functions and simultaneously learns stripe features in the angular direction of the sinogram data. These two components are combined within an optimization constraint framework, achieving unsupervised iterative correction of stripe artifacts in the projection domain. Experimental results demonstrate that the proposed method significantly outperforms current state-of-the-art techniques in removing ring artifacts while maintaining the clarity of CT images., Comment: 10 pages, 11 figures

Current video summarization methods largely rely on transformer-based architectures, which, due to their quadratic complexity, require substantial computational resources. In this work, we address these inefficiencies by enhancing the Direct-to-Summarize Network (DSNet) with more resource-efficient token mixing mechanisms. We show that replacing traditional attention with alternatives like Fourier, Wavelet transforms, and Nystr\"omformer improves efficiency and performance. Furthermore, we explore various pooling strategies within the Regional Proposal Network, including ROI pooling, Fast Fourier Transform pooling, and flat pooling. Our experimental results on TVSum and SumMe datasets demonstrate that these modifications significantly reduce computational costs while maintaining competitive summarization performance. Thus, our work offers a more scalable solution for video summarization tasks., Comment: 10 pages, 5 figures

Achieving human-like memory recall in artificial systems remains a challenging frontier in computer vision. Humans demonstrate remarkable ability to recall images after a single exposure, even after being shown thousands of images. However, this capacity diminishes significantly when confronted with non-natural stimuli such as random textures. In this paper, we present a method inspired by human memory processes to bridge this gap between artificial and biological memory systems. Our approach focuses on encoding images to mimic the high-level information retained by the human brain, rather than storing raw pixel data. By adding noise to images before encoding, we introduce variability akin to the non-deterministic nature of human memory encoding. Leveraging pre-trained models' embedding layers, we explore how different architectures encode images and their impact on memory recall. Our method achieves impressive results, with 97% accuracy on natural images and near-random performance (52%) on textures. We provide insights into the encoding process and its implications for machine learning memory systems, shedding light on the parallels between human and artificial intelligence memory mechanisms., Comment: 5 pages, 7 figures

Recent advancements in single image super-resolution have been predominantly driven by token mixers and transformer architectures. WaveMixSR utilized the WaveMix architecture, employing a two-dimensional discrete wavelet transform for spatial token mixing, achieving superior performance in super-resolution tasks with remarkable resource efficiency. In this work, we present an enhanced version of the WaveMixSR architecture by (1) replacing the traditional transpose convolution layer with a pixel shuffle operation and (2) implementing a multistage design for higher resolution tasks ($4\times$). Our experiments demonstrate that our enhanced model -- WaveMixSR-V2 -- outperforms other architectures in multiple super-resolution tasks, achieving state-of-the-art for the BSD100 dataset, while also consuming fewer resources, exhibits higher parameter efficiency, lower latency and higher throughput. Our code is available at https://github.com/pranavphoenix/WaveMixSR., Comment: 10 pages. Accepted in AAAI 2025. arXiv admin note: text overlap with arXiv:2307.00430

Learning semantically meaningful image transformations (i.e. rotation, thickness, blur) directly from examples can be a challenging task. Recently, the Manifold Autoencoder (MAE) proposed using a set of Lie group operators to learn image transformations directly from examples. However, this approach has limitations, as the learned operators are not guaranteed to be disentangled and the training routine is prohibitively expensive when scaling up the model. To address these limitations, we propose MANGO (transformation Manifolds with Grouped Operators) for learning disentangled operators that describe image transformations in distinct latent subspaces. Moreover, our approach allows practitioners the ability to define which transformations they aim to model, thus improving the semantic meaning of the learned operators. Through our experiments, we demonstrate that MANGO enables composition of image transformations and introduces a one-phase training routine that leads to a 100x speedup over prior works., Comment: Submitted to IEEE ICASSP 2025. This work has been submitted to the IEEE for possible publication

In the realm of image synthesis, achieving fidelity to a reference image while adhering to conditional prompts remains a significant challenge. This paper proposes a novel approach that integrates a diffusion model with latent space manipulation and gradient-based selective attention mechanisms to address this issue. Leveraging Grad-SAM (Gradient-based Selective Attention Manipulation), we analyze the cross attention maps of the cross attention layers and gradients for the denoised latent vector, deriving importance scores of elements of denoised latent vector related to the subject of interest. Using this information, we create masks at specific timesteps during denoising to preserve subjects while seamlessly integrating the reference image features. This approach ensures the faithful formation of subjects based on conditional prompts, while concurrently refining the background for a more coherent composition. Our experiments on places365 dataset demonstrate promising results, with our proposed model achieving the lowest mean and median Frechet Inception Distance (FID) scores compared to baseline models, indicating superior fidelity preservation. Furthermore, our model exhibits competitive performance in aligning the generated images with provided textual descriptions, as evidenced by high CLIP scores. These results highlight the effectiveness of our approach in both fidelity preservation and textual context preservation, offering a significant advancement in text-to-image synthesis tasks., Comment: 7 pages , 5 figures

We give an overview of combinatorial methods to represent 3D data, such as graphs and meshes, from the viewpoint of their amenability to analysis using machine learning algorithms. We highlight pros and cons of various representations and we discuss some methods of generating/switching between the representations. We finally present two concrete applications in life science and industry. Despite its theoretical nature, our discussion is in general motivated by, and biased towards real-world challenges., Comment: 14 pages, 11 figures

Vision Transformer (ViT) architectures traditionally employ a grid-based approach to tokenization independent of the semantic content of an image. We propose a modular superpixel tokenization strategy which decouples tokenization and feature extraction; a shift from contemporary approaches where these are treated as an undifferentiated whole. Using on-line content-aware tokenization and scale- and shape-invariant positional embeddings, we perform experiments and ablations that contrast our approach with patch-based tokenization and randomized partitions as baselines. We show that our method significantly improves the faithfulness of attributions, gives pixel-level granularity on zero-shot unsupervised dense prediction tasks, while maintaining predictive performance in classification tasks. Our approach provides a modular tokenization framework commensurable with standard architectures, extending the space of ViTs to a larger class of semantically-rich models., Comment: To appear in ECCV (MELEX) 2024 Workshop Proceedings

The generative self-supervised learning strategy exhibits remarkable learning representational capabilities. However, there is limited attention to end-to-end pre-training methods based on a hybrid architecture of CNN and Transformer, which can learn strong local and global representations simultaneously. To address this issue, we propose a generative pre-training strategy called Hybrid Sparse masKing (HySparK) based on masked image modeling and apply it to large-scale pre-training on medical images. First, we perform a bottom-up 3D hybrid masking strategy on the encoder to keep consistency masking. Then we utilize sparse convolution for the top CNNs and encode unmasked patches for the bottom vision Transformers. Second, we employ a simple hierarchical decoder with skip-connections to achieve dense multi-scale feature reconstruction. Third, we implement our pre-training method on a collection of multiple large-scale 3D medical imaging datasets. Extensive experiments indicate that our proposed pre-training strategy demonstrates robust transfer-ability in supervised downstream tasks and sheds light on HySparK's promising prospects. The code is available at https://github.com/FengheTan9/HySparK, Comment: Early accept at MICCAI 2024

Human motion prediction based on 3D skeleton is a significant challenge in computer vision, primarily focusing on the effective representation of motion. In this paper, we propose a self-supervised learning framework designed to enhance motion representation. This framework consists of two stages: first, the network is pretrained through the self-reconstruction of past sequences, and the guided reconstruction of future sequences based on past movements. We design a velocity-based mask strategy to focus on the joints with large-scale moving. Subsequently, the pretrained network undergoes finetuning for specific tasks. Self-reconstruction, guided by patterns of past motion, substantially improves the model's ability to represent the spatiotemporal relationships among joints but also captures the latent relationships between past and future sequences. This capability is crucial for motion prediction tasks that solely depend on historical motion data. By employing this straightforward yet effective training paradigm, our method outperforms existing \textit{state-of-the-art} methods, reducing the average prediction errors by 8.8\% across Human3.6M, 3DPW, and AMASS datasets. The code is available at https://github.com/JunyuShi02/PMG-MRL., Comment: 13 pages, 4 figures

Deep learning has revolutionized medical image segmentation, but it relies heavily on high-quality annotations. The time, cost and expertise required to label images at the pixel-level for each new task has slowed down widespread adoption of the paradigm. We propose Pix2Rep, a self-supervised learning (SSL) approach for few-shot segmentation, that reduces the manual annotation burden by learning powerful pixel-level representations directly from unlabeled images. Pix2Rep is a novel pixel-level loss and pre-training paradigm for contrastive SSL on whole images. It is applied to generic encoder-decoder deep learning backbones (e.g., U-Net). Whereas most SSL methods enforce invariance of the learned image-level representations under intensity and spatial image augmentations, Pix2Rep enforces equivariance of the pixel-level representations. We demonstrate the framework on a task of cardiac MRI segmentation. Results show improved performance compared to existing semi- and self-supervised approaches; and a 5-fold reduction in the annotation burden for equivalent performance versus a fully supervised U-Net baseline. This includes a 30% (resp. 31%) DICE improvement for one-shot segmentation under linear-probing (resp. fine-tuning). Finally, we also integrate the novel Pix2Rep concept with the Barlow Twins non-contrastive SSL, which leads to even better segmentation performance., Comment: Accepted at MIDL 2024

This study proposes a neural disparity field (NDF) that establishes an implicit, continuous representation of scene disparity based on a neural field and an iterative approach to address the inverse problem of NDF reconstruction from light-field data. NDF enables seamless and precise characterization of disparity variations in three-dimensional scenes and can discretize disparity at any arbitrary resolution, overcoming the limitations of traditional disparity maps that are prone to sampling errors and interpolation inaccuracies. The proposed NDF network architecture utilizes hash encoding combined with multilayer perceptrons to capture detailed disparities in texture levels, thereby enhancing its ability to represent the geometric information of complex scenes. By leveraging the spatial-angular consistency inherent in light-field data, a differentiable forward model to generate a central view image from the light-field data is developed. Based on the forward model, an optimization scheme for the inverse problem of NDF reconstruction using differentiable propagation operators is established. Furthermore, an iterative solution method is adopted to reconstruct the NDF in the optimization scheme, which does not require training datasets and applies to light-field data captured by various acquisition methods. Experimental results demonstrate that high-quality NDF can be reconstructed from light-field data using the proposed method. High-resolution disparity can be effectively recovered by NDF, demonstrating its capability for the implicit, continuous representation of scene disparities., Comment: 12 pages, 7 figures

Federated Learning (FL) offers a privacy-preserving approach to train models on decentralized data. Its potential in healthcare is significant, but challenges arise due to cross-client variations in medical image data, exacerbated by limited annotations. This paper introduces Cross-Client Variations Adaptive Federated Learning (CCVA-FL) to address these issues. CCVA-FL aims to minimize cross-client variations by transforming images into a common feature space. It involves expert annotation of a subset of images from each client, followed by the selection of a client with the least data complexity as the target. Synthetic medical images are then generated using Scalable Diffusion Models with Transformers (DiT) based on the target client's annotated images. These synthetic images, capturing diversity and representing the original data, are shared with other clients. Each client then translates its local images into the target image space using image-to-image translation. The translated images are subsequently used in a federated learning setting to develop a server model. Our results demonstrate that CCVA-FL outperforms Vanilla Federated Averaging by effectively addressing data distribution differences across clients without compromising privacy., Comment: I found critical errors in the manuscript affecting its validity. I need to correct these before resubmitting. Major changes to methodology and results are underway, significantly altering the content. I will resubmit the revised version

The heterogeneity of neurological conditions, ranging from structural anomalies to functional impairments, presents a significant challenge in medical imaging analysis tasks. Moreover, the limited availability of well-annotated datasets constrains the development of robust analysis models. Against this backdrop, this study introduces a novel approach leveraging the inherent anatomical symmetrical features of the human brain to enhance the subsequent detection and segmentation analysis for brain diseases. A novel Symmetry-Aware Cross-Attention (SACA) module is proposed to encode symmetrical features of left and right hemispheres, and a proxy task to detect symmetrical features as the Symmetry-Aware Head (SAH) is proposed, which guides the pretraining of the whole network on a vast 3D brain imaging dataset comprising both healthy and diseased brain images across various MRI and CT. Through meticulous experimentation on downstream tasks, including both classification and segmentation for brain diseases, our model demonstrates superior performance over state-of-the-art methodologies, particularly highlighting the significance of symmetry-aware learning. Our findings advocate for the effectiveness of incorporating symmetry awareness into pretraining and set a new benchmark for medical imaging analysis, promising significant strides toward accurate and efficient diagnostic processes. Code is available at https://github.com/bitMyron/sa-swin., Comment: MICCAI 2024

The robustness of machine learning models has been questioned by the existence of adversarial examples. We examine the threat of adversarial examples in practical applications that require lightweight models for one-class classification. Building on Ilyas et al. (2019), we investigate the vulnerability of lightweight one-class classifiers to adversarial attacks and possible reasons for it. Our results show that lightweight one-class classifiers learn features that are not robust (e.g. texture) under stronger attacks. However, unlike in multi-class classification (Ilyas et al., 2019), these non-robust features are not always useful for the one-class task, suggesting that learning these unpredictive and non-robust features is an unwanted consequence of training., Comment: CVPR Visual and Anomaly Detection (VAND) Workshop 2024

Advances in generative models have led to significant interest in image synthesis, demonstrating the ability to generate high-quality images for a diverse range of text prompts. Despite this progress, most studies ignore the presence of bias. In this paper, we examine several text-to-image models not only by qualitatively assessing their performance in generating accurate images of human faces, groups, and specified numbers of objects but also by presenting a social bias analysis. As expected, models with larger capacity generate higher-quality images. However, we also document the inherent gender or social biases these models possess, offering a more complete understanding of their impact and limitations., Comment: 20 pages, 8 figures

In this paper, we take a new approach to autoregressive image generation that is based on two main ingredients. The first is wavelet image coding, which allows to tokenize the visual details of an image from coarse to fine details by ordering the information starting with the most significant bits of the most significant wavelet coefficients. The second is a variant of a language transformer whose architecture is re-designed and optimized for token sequences in this 'wavelet language'. The transformer learns the significant statistical correlations within a token sequence, which are the manifestations of well-known correlations between the wavelet subbands at various resolutions. We show experimental results with conditioning on the generation process., Comment: 17 pages, 11 figures

The European Space Agency's Copernicus Sentinel-1 (S-1) mission is a constellation of C-band synthetic aperture radar (SAR) satellites that provide unprecedented monitoring of the world's oceans. S-1's wave mode (WV) captures 20x20 km image patches at 5 m pixel resolution and is unaffected by cloud cover or time-of-day. The mission's open data policy has made SAR data easily accessible for a range of applications, but the need for manual image annotations is a bottleneck that hinders the use of machine learning methods. This study uses nearly 10 million WV-mode images and contrastive self-supervised learning to train a semantic embedding model called WV-Net. In multiple downstream tasks, WV-Net outperforms a comparable model that was pre-trained on natural images (ImageNet) with supervised learning. Experiments show improvements for estimating wave height (0.50 vs 0.60 RMSE using linear probing), estimating near-surface air temperature (0.90 vs 0.97 RMSE), and performing multilabel-classification of geophysical and atmospheric phenomena (0.96 vs 0.95 micro-averaged AUROC). WV-Net embeddings are also superior in an unsupervised image-retrieval task and scale better in data-sparse settings. Together, these results demonstrate that WV-Net embeddings can support geophysical research by providing a convenient foundation model for a variety of data analysis and exploration tasks., Comment: 20 pages, 9 figures, submitted to NeurIPS 2024

Histopathological analysis of Whole Slide Images (WSIs) has seen a surge in the utilization of deep learning methods, particularly Convolutional Neural Networks (CNNs). However, CNNs often fall short in capturing the intricate spatial dependencies inherent in WSIs. Graph Neural Networks (GNNs) present a promising alternative, adept at directly modeling pairwise interactions and effectively discerning the topological tissue and cellular structures within WSIs. Recognizing the pressing need for deep learning techniques that harness the topological structure of WSIs, the application of GNNs in histopathology has experienced rapid growth. In this comprehensive review, we survey GNNs in histopathology, discuss their applications, and explore emerging trends that pave the way for future advancements in the field. We begin by elucidating the fundamentals of GNNs and their potential applications in histopathology. Leveraging quantitative literature analysis, we identify four emerging trends: Hierarchical GNNs, Adaptive Graph Structure Learning, Multimodal GNNs, and Higher-order GNNs. Through an in-depth exploration of these trends, we offer insights into the evolving landscape of GNNs in histopathological analysis. Based on our findings, we propose future directions to propel the field forward. Our analysis serves to guide researchers and practitioners towards innovative approaches and methodologies, fostering advancements in histopathological analysis through the lens of graph neural networks.

Recent advances in multi-modal algorithms have driven and been driven by the increasing availability of large image-text datasets, leading to significant strides in various fields, including computational pathology. However, in most existing medical image-text datasets, the text typically provides high-level summaries that may not sufficiently describe sub-tile regions within a large pathology image. For example, an image might cover an extensive tissue area containing cancerous and healthy regions, but the accompanying text might only specify that this image is a cancer slide, lacking the nuanced details needed for in-depth analysis. In this study, we introduce STimage-1K4M, a novel dataset designed to bridge this gap by providing genomic features for sub-tile images. STimage-1K4M contains 1,149 images derived from spatial transcriptomics data, which captures gene expression information at the level of individual spatial spots within a pathology image. Specifically, each image in the dataset is broken down into smaller sub-image tiles, with each tile paired with 15,000-30,000 dimensional gene expressions. With 4,293,195 pairs of sub-tile images and gene expressions, STimage-1K4M offers unprecedented granularity, paving the way for a wide range of advanced research in multi-modal data analysis an innovative applications in computational pathology, and beyond.

In contemporary computer vision applications, particularly image classification, architectural backbones pre-trained on large datasets like ImageNet are commonly employed as feature extractors. Despite the widespread use of these pre-trained convolutional neural networks (CNNs), there remains a gap in understanding the performance of various resource-efficient backbones across diverse domains and dataset sizes. Our study systematically evaluates multiple lightweight, pre-trained CNN backbones under consistent training settings across a variety of datasets, including natural images, medical images, galaxy images, and remote sensing images. This comprehensive analysis aims to aid machine learning practitioners in selecting the most suitable backbone for their specific problem, especially in scenarios involving small datasets where fine-tuning a pre-trained network is crucial. Even though attention-based architectures are gaining popularity, we observed that they tend to perform poorly under low data finetuning tasks compared to CNNs. We also observed that some CNN architectures such as ConvNeXt, RegNet and EfficientNet performs well compared to others on a diverse set of domains consistently. Our findings provide actionable insights into the performance trade-offs and effectiveness of different backbones, facilitating informed decision-making in model selection for a broad spectrum of computer vision domains. Our code is available here: https://github.com/pranavphoenix/Backbones, Comment: 12 pages, 2 figures

Cuneiform tablets, emerging in ancient Mesopotamia around the late fourth millennium BCE, represent one of humanity's earliest writing systems. Characterized by wedge-shaped marks on clay tablets, these artifacts provided insight into Mesopotamian civilization across various domains. Traditionally, the analysis and dating of these tablets rely on subjective assessment of shape and writing style, leading to uncertainties in pinpointing their exact temporal origins. Recent advances in digitization have revolutionized the study of cuneiform by enhancing accessibility and analytical capabilities. Our research uniquely focuses on the silhouette of tablets as significant indicators of their historical periods, diverging from most studies that concentrate on textual content. Utilizing an unprecedented dataset of over 94,000 images from the Cuneiform Digital Library Initiative collection, we apply deep learning methods to classify cuneiform tablets, covering over 3,000 years of history. By leveraging statistical, computational techniques, and generative modeling through Variational Auto-Encoders (VAEs), we achieve substantial advancements in the automatic classification of these ancient documents, focusing on the tablets' silhouettes as key predictors. Our classification approach begins with a Decision Tree using height-to-width ratios and culminates with a ResNet50 model, achieving a 61% macro F1-score for tablet silhouettes. Moreover, we introduce novel VAE-powered tools to enhance explainability and enable researchers to explore changes in tablet shapes across different eras and genres. This research contributes to document analysis and diplomatics by demonstrating the value of large-scale data analysis combined with statistical methods. These insights offer valuable tools for historians and epigraphists, enriching our understanding of cuneiform tablets and the cultures that produced them., Comment: 24 pages, 18 figures

It has recently been discovered that using a pre-trained vision-language model (VLM), e.g., CLIP, to align a whole query image with several finer text descriptions generated by a large language model can significantly enhance zero-shot performance. However, in this paper, we empirically find that the finer descriptions tend to align more effectively with local areas of the query image rather than the whole image, and then we theoretically validate this finding. Thus, we present a method called weighted visual-text cross alignment (WCA). This method begins with a localized visual prompting technique, designed to identify local visual areas within the query image. The local visual areas are then cross-aligned with the finer descriptions by creating a similarity matrix using the pre-trained VLM. To determine how well a query image aligns with each category, we develop a score function based on the weighted similarities in this matrix. Extensive experiments demonstrate that our method significantly improves zero-shot performance across various datasets, achieving results that are even comparable to few-shot learning methods., Comment: 22 pages, 16 figures, published to ICML 2024

Image-based 3D reconstruction is a challenging task that involves inferring the 3D shape of an object or scene from a set of input images. Learning-based methods have gained attention for their ability to directly estimate 3D shapes. This review paper focuses on state-of-the-art techniques for 3D reconstruction, including the generation of novel, unseen views. An overview of recent developments in the Gaussian Splatting method is provided, covering input types, model structures, output representations, and training strategies. Unresolved challenges and future directions are also discussed. Given the rapid progress in this domain and the numerous opportunities for enhancing 3D reconstruction methods, a comprehensive examination of algorithms appears essential. Consequently, this study offers a thorough overview of the latest advancements in Gaussian Splatting., Comment: 24 pages

Generalization to unseen data remains poorly understood for deep learning classification and foundation models, especially in the open set scenario. How can one assess the ability of networks to adapt to new or extended versions of their input space in the spirit of few-shot learning, out-of-distribution generalization, domain adaptation, and category discovery? Which layers of a network are likely to generalize best? We provide a new method for evaluating the capacity of networks to represent a sampled domain, regardless of whether the network has been trained on all classes in that domain. Our approach is the following: after fine-tuning state-of-the-art pre-trained models for visual classification on a particular domain, we assess their performance on data from related but distinct variations in that domain. Generalization power is quantified as a function of the latent embeddings of unseen data from intermediate layers for both unsupervised and supervised settings. Working throughout all stages of the network, we find that (i) high classification accuracy does not imply high generalizability; and (ii) deeper layers in a model do not always generalize the best, which has implications for pruning. Since the trends observed across datasets are largely consistent, we conclude that our approach reveals (a function of) the intrinsic capacity of the different layers of a model to generalize. Our code is available at https://github.com/dyballa/generalization, Comment: 7 pages, 6 figures

Recent dataset deduplication techniques have demonstrated that content-aware dataset pruning can dramatically reduce the cost of training Vision-Language Pretrained (VLP) models without significant performance losses compared to training on the original dataset. These results have been based on pruning commonly used image-caption datasets collected from the web -- datasets that are known to harbor harmful social biases that may then be codified in trained models. In this work, we evaluate how deduplication affects the prevalence of these biases in the resulting trained models and introduce an easy-to-implement modification to the recent SemDeDup algorithm that can reduce the negative effects that we observe. When examining CLIP-style models trained on deduplicated variants of LAION-400M, we find our proposed FairDeDup algorithm consistently leads to improved fairness metrics over SemDeDup on the FairFace and FACET datasets while maintaining zero-shot performance on CLIP benchmarks., Comment: Conference paper at CVPR 2024. 6 pages, 8 figures. Project Page: https://ericslyman.com/fairdedup/

Degraded underwater images decrease the accuracy of underwater object detection. However, existing methods for underwater image enhancement mainly focus on improving the indicators in visual aspects, which may not benefit the tasks of underwater image detection, and may lead to serious degradation in performance. To alleviate this problem, we proposed a bidirectional-guided method for underwater object detection, referred to as BG-YOLO. In the proposed method, network is organized by constructing an enhancement branch and a detection branch in a parallel way. The enhancement branch consists of a cascade of an image enhancement subnet and an object detection subnet. And the detection branch only consists of a detection subnet. A feature guided module connects the shallow convolution layer of the two branches. When training the enhancement branch, the object detection subnet in the enhancement branch guides the image enhancement subnet to be optimized towards the direction that is most conducive to the detection task. The shallow feature map of the trained enhancement branch will be output to the feature guided module, constraining the optimization of detection branch through consistency loss and prompting detection branch to learn more detailed information of the objects. And hence the detection performance will be refined. During the detection tasks, only detection branch will be reserved so that no additional cost of computation will be introduced. Extensive experiments demonstrate that the proposed method shows significant improvement in performance of the detector in severely degraded underwater scenes while maintaining a remarkable detection speed., Comment: 15 pages, 8 figures, 4 tables

In spectral CT reconstruction, the basis materials decomposition involves solving a large-scale nonlinear system of integral equations, which is highly ill-posed mathematically. This paper proposes a model that parameterizes the attenuation coefficients of the object using a neural field representation, thereby avoiding the complex calculations of pixel-driven projection coefficient matrices during the discretization process of line integrals. It introduces a lightweight discretization method for line integrals based on a ray-driven neural field, enhancing the accuracy of the integral approximation during the discretization process. The basis materials are represented as continuous vector-valued implicit functions to establish a neural field parameterization model for the basis materials. The auto-differentiation framework of deep learning is then used to solve the implicit continuous function of the neural base-material fields. This method is not limited by the spatial resolution of reconstructed images, and the network has compact and regular properties. Experimental validation shows that our method performs exceptionally well in addressing the spectral CT reconstruction. Additionally, it fulfils the requirements for the generation of high-resolution reconstruction images., Comment: 14 pages,16 figures

We propose a novel architecture and method of explainable classification with Concept Bottleneck Models (CBMs). While SOTA approaches to Image Classification task work as a black box, there is a growing demand for models that would provide interpreted results. Such a models often learn to predict the distribution over class labels using additional description of this target instances, called concepts. However, existing Bottleneck methods have a number of limitations: their accuracy is lower than that of a standard model and CBMs require an additional set of concepts to leverage. We provide a framework for creating Concept Bottleneck Model from pre-trained multi-modal encoder and new CLIP-like architectures. By introducing a new type of layers known as Concept Bottleneck Layers, we outline three methods for training them: with $\ell_1$-loss, contrastive loss and loss function based on Gumbel-Softmax distribution (Sparse-CBM), while final FC layer is still trained with Cross-Entropy. We show a significant increase in accuracy using sparse hidden layers in CLIP-based bottleneck models. Which means that sparse representation of concepts activation vector is meaningful in Concept Bottleneck Models. Moreover, with our Concept Matrix Search algorithm we can improve CLIP predictions on complex datasets without any additional training or fine-tuning. The code is available at: https://github.com/Andron00e/SparseCBM., Comment: 23 pages, 1 algorithm, 36 figures

Laparoscopic video tracking primarily focuses on two target types: surgical instruments and anatomy. The former could be used for skill assessment, while the latter is necessary for the projection of virtual overlays. Where instrument and anatomy tracking have often been considered two separate problems, in this paper, we propose a method for joint tracking of all structures simultaneously. Based on a single 2D monocular video clip, we train a neural field to represent a continuous spatiotemporal scene, used to create 3D tracks of all surfaces visible in at least one frame. Due to the small size of instruments, they generally cover a small part of the image only, resulting in decreased tracking accuracy. Therefore, we propose enhanced class weighting to improve the instrument tracks. We evaluate tracking on video clips from laparoscopic cholecystectomies, where we find mean tracking accuracies of 92.4% for anatomical structures and 87.4% for instruments. Additionally, we assess the quality of depth maps obtained from the method's scene reconstructions. We show that these pseudo-depths have comparable quality to a state-of-the-art pre-trained depth estimator. On laparoscopic videos in the SCARED dataset, the method predicts depth with an MAE of 2.9 mm and a relative error of 9.2%. These results show the feasibility of using neural fields for monocular 3D reconstruction of laparoscopic scenes.

We introduce TexTile, a novel differentiable metric to quantify the degree upon which a texture image can be concatenated with itself without introducing repeating artifacts (i.e., the tileability). Existing methods for tileable texture synthesis focus on general texture quality, but lack explicit analysis of the intrinsic repeatability properties of a texture. In contrast, our TexTile metric effectively evaluates the tileable properties of a texture, opening the door to more informed synthesis and analysis of tileable textures. Under the hood, TexTile is formulated as a binary classifier carefully built from a large dataset of textures of different styles, semantics, regularities, and human annotations.Key to our method is a set of architectural modifications to baseline pre-train image classifiers to overcome their shortcomings at measuring tileability, along with a custom data augmentation and training regime aimed at increasing robustness and accuracy. We demonstrate that TexTile can be plugged into different state-of-the-art texture synthesis methods, including diffusion-based strategies, and generate tileable textures while keeping or even improving the overall texture quality. Furthermore, we show that TexTile can objectively evaluate any tileable texture synthesis method, whereas the current mix of existing metrics produces uncorrelated scores which heavily hinders progress in the field., Comment: CVPR 2024. Project page: https://mslab.es/projects/TexTile/

As part of human core knowledge, the representation of objects is the building block of mental representation that supports high-level concepts and symbolic reasoning. While humans develop the ability of perceiving objects situated in 3D environments without supervision, models that learn the same set of abilities with similar constraints faced by human infants are lacking. Towards this end, we developed a novel network architecture that simultaneously learns to 1) segment objects from discrete images, 2) infer their 3D locations, and 3) perceive depth, all while using only information directly available to the brain as training data, namely: sequences of images and self-motion. The core idea is treating objects as latent causes of visual input which the brain uses to make efficient predictions of future scenes. This results in object representations being learned as an essential byproduct of learning to predict., Comment: 21 pages, 11 figures. Project webpage can be found at https://jday54.github.io/opple_site/

Dimensionality reduction can be applied to hyperspectral images so that the most useful data can be extracted and processed more quickly. This is critical in any situation in which data volume exceeds the capacity of the computational resources, particularly in the case of remote sensing platforms (e.g., drones, satellites), but also in the case of multi-year datasets. Moreover, the computational strategies of unsupervised dimensionality reduction often provide the basis for more complicated supervised techniques. Seven unsupervised dimensionality reduction algorithms are tested on hyperspectral data from the HYPSO-1 earth observation satellite. Each particular algorithm is chosen to be representative of a broader collection. The experiments probe the computational complexity, reconstruction accuracy, signal clarity, sensitivity to artifacts, and effects on target detection and classification of the different algorithms. No algorithm consistently outperformed the others across all tests, but some general trends regarding the characteristics of the algorithms did emerge. With half a million pixels, computational time requirements of the methods varied by 5 orders of magnitude, and the reconstruction error varied by about 3 orders of magnitude. A relationship between mutual information and artifact susceptibility was suggested by the tests. The relative performance of the algorithms differed significantly between the target detection and classification tests. Overall, these experiments both show the power of dimensionality reduction and give guidance regarding how to evaluate a technique prior to incorporating it into a processing pipeline., Comment: 16 pages, 11 figures

In an unfamiliar setting, a model-based reinforcement learning agent can be limited by the accuracy of its world model. In this work, we present a novel, training-free approach to improving the performance of such agents separately from planning and learning. We do so by applying iterative inference at decision-time, to fine-tune the inferred agent states based on the coherence of future state representations. Our approach achieves a consistent improvement in both reconstruction accuracy and task performance when applied to visual 3D navigation tasks. We go on to show that considering more future states further improves the performance of the agent in partially-observable environments, but not in a fully-observable one. Finally, we demonstrate that agents with less training pre-evaluation benefit most from our approach.

Generalization to unseen data is a key desideratum for deep networks, but its relation to classification accuracy is unclear. Using a minimalist vision dataset and a measure of generalizability, we show that popular networks, from deep convolutional networks (CNNs) to transformers, vary in their power to extrapolate to unseen classes both across layers and across architectures. Accuracy is not a good predictor of generalizability, and generalization varies non-monotonically with layer depth., Comment: Accepted as a Tiny Paper at ICLR 2024. Code available at https://github.com/dyballa/generalization/tree/ICLR2024TinyPaper

In recent years, there has been a growing interest in training Neural Networks to approximate Unsigned Distance Fields (UDFs) for representing open surfaces in the context of 3D reconstruction. However, UDFs are non-differentiable at the zero level set which leads to significant errors in distances and gradients, generally resulting in fragmented and discontinuous surfaces. In this paper, we propose to learn a hyperbolic scaling of the unsigned distance field, which defines a new Eikonal problem with distinct boundary conditions. This allows our formulation to integrate seamlessly with state-of-the-art continuously differentiable implicit neural representation networks, largely applied in the literature to represent signed distance fields. Our approach not only addresses the challenge of open surface representation but also demonstrates significant improvement in reconstruction quality and training performance. Moreover, the unlocked field's differentiability allows the accurate computation of essential topological properties such as normal directions and curvatures, pervasive in downstream tasks such as rendering. Through extensive experiments, we validate our approach across various data sets and against competitive baselines. The results demonstrate enhanced accuracy and up to an order of magnitude increase in speed compared to previous methods.

We seek to enable classic processing of continuous ultra-sparse spatiotemporal data generated by event-based sensors with dense machine learning models. We propose a novel hybrid pipeline composed of asynchronous sensing and synchronous processing that combines several ideas: (1) an embedding based on PointNet models -- the ALERT module -- that can continuously integrate new and dismiss old events thanks to a leakage mechanism, (2) a flexible readout of the embedded data that allows to feed any downstream model with always up-to-date features at any sampling rate, (3) exploiting the input sparsity in a patch-based approach inspired by Vision Transformer to optimize the efficiency of the method. These embeddings are then processed by a transformer model trained for object and gesture recognition. Using this approach, we achieve performances at the state-of-the-art with a lower latency than competitors. We also demonstrate that our asynchronous model can operate at any desired sampling rate., Comment: Originally published in the Proceedings of Machine Learning Research ICML 2024

Recently proposed methods for implicitly representing signals such as images, scenes, or geometries using coordinate-based neural network architectures often do not leverage the choice of activation functions, or do so only to a limited extent. In this paper, we introduce the Hyperbolic Oscillation function (HOSC), a novel activation function with a controllable sharpness parameter. Unlike any previous activations, HOSC has been specifically designed to better capture sudden changes in the input signal, and hence sharp or acute features of the underlying data, as well as smooth low-frequency transitions. Due to its simplicity and modularity, HOSC offers a plug-and-play functionality that can be easily incorporated into any existing method employing a neural network as a way of implicitly representing a signal. We benchmark HOSC against other popular activations in an array of general tasks, empirically showing an improvement in the quality of obtained representations, provide the mathematical motivation behind the efficacy of HOSC, and discuss its limitations., Comment: 12 pages, 7 figures