Computed tomography (CT) and clinical numeric data are essential modalities for cancer evaluation, but building large-scale multimodal training datasets for developing medical foundation models remains challenging due to the structural complexity of multi-slice CT data and high cost of expert annotation. In this study, we propose MEDFORM, a multimodal pre-training strategy that guides CT image representation learning using complementary information from clinical data for medical foundation model development. MEDFORM efficiently processes CT slice through multiple instance learning (MIL) and adopts a dual pre-training strategy: first pretraining the CT slice feature extractor using SimCLR-based self-supervised learning, then aligning CT and clinical modalities through cross-modal contrastive learning. Our model was pre-trained on three different cancer types: lung cancer (141,171 slices), breast cancer (8,100 slices), colorectal cancer (10,393 slices). The experimental results demonstrated that this dual pre-training strategy improves cancer classification performance and maintains robust performance in few-shot learning scenarios. Code available at https://github.com/DigitalHealthcareLab/25MultiModalFoundationModel.git
Analyzing a player's technique in table tennis requires knowledge of the ball's 3D trajectory and spin. While, the spin is not directly observable in standard broadcasting videos, we show that it can be inferred from the ball's trajectory in the video. We present a novel method to infer the initial spin and 3D trajectory from the corresponding 2D trajectory in a video. Without ground truth labels for broadcast videos, we train a neural network solely on synthetic data. Due to the choice of our input data representation, physically correct synthetic training data, and using targeted augmentations, the network naturally generalizes to real data. Notably, these simple techniques are sufficient to achieve generalization. No real data at all is required for training. To the best of our knowledge, we are the first to present a method for spin and trajectory prediction in simple monocular broadcast videos, achieving an accuracy of 92.0% in spin classification and a 2D reprojection error of 0.19% of the image diagonal.
Lifelong learning algorithms enable models to incrementally acquire new knowledge without forgetting previously learned information. Contrarily, the field of machine unlearning focuses on explicitly forgetting certain previous knowledge from pretrained models when requested, in order to comply with data privacy regulations on the right-to-be-forgotten. Enabling efficient lifelong learning with the capability to selectively unlearn sensitive information from models presents a critical and largely unaddressed challenge with contradicting objectives. We address this problem from the perspective of simultaneously preventing catastrophic forgetting and allowing forward knowledge transfer during task-incremental learning, while ensuring exact task unlearning and minimizing memory requirements, based on a single neural network model to be adapted. Our proposed solution, privacy-aware lifelong learning (PALL), involves optimization of task-specific sparse subnetworks with parameter sharing within a single architecture. We additionally utilize an episodic memory rehearsal mechanism to facilitate exact unlearning without performance degradations. We empirically demonstrate the scalability of PALL across various architectures in image classification, and provide a state-of-the-art solution that uniquely integrates lifelong learning and privacy-aware unlearning mechanisms for responsible AI applications.
Automated design of neural network architectures tailored for a specific task is an extremely promising, albeit inherently difficult, avenue to explore. While most results in this domain have been achieved on image classification and language modelling problems, here we concentrate on dense per-pixel tasks, in particular, semantic image segmentation using fully convolutional networks. In contrast to the aforementioned areas, the design choices of a fully convolutional network require several changes, ranging from the sort of operations that need to be used---e.g., dilated convolutions---to a solving of a more difficult optimisation problem. In this work, we are particularly interested in searching for high-performance compact segmentation architectures, able to run in real-time using limited resources. To achieve that, we intentionally over-parameterise the architecture during the training time via a set of auxiliary cells that provide an intermediate supervisory signal and can be omitted during the evaluation phase. The design of the auxiliary cell is emitted by a controller, a neural network with the fixed structure trained using reinforcement learning. More crucially, we demonstrate how to efficiently search for these architectures within limited time and computational budgets. In particular, we rely on a progressive strategy that terminates non-promising architectures from being further trained, and on Polyak averaging coupled with knowledge distillation to speed-up the convergence. Quantitatively, in 8 GPU-days our approach discovers a set of architectures performing on-par with state-of-the-art among compact models on the semantic segmentation, pose estimation and depth prediction tasks. Code will be made available here: https://github.com/drsleep/nas-segm-pytorch
Clinicians are often very sceptical about applying automatic image processing approaches, especially deep learning based methods, in practice. One main reason for this is the black-box nature of these approaches and the inherent problem of missing insights of the automatically derived decisions. In order to increase trust in these methods, this paper presents approaches that help to interpret and explain the results of deep learning algorithms by depicting the anatomical areas which influence the decision of the algorithm most. Moreover, this research presents a unified framework, TorchEsegeta, for applying various interpretability and explainability techniques for deep learning models and generate visual interpretations and explanations for clinicians to corroborate their clinical findings. In addition, this will aid in gaining confidence in such methods. The framework builds on existing interpretability and explainability techniques that are currently focusing on classification models, extending them to segmentation tasks. In addition, these methods have been adapted to 3D models for volumetric analysis. The proposed framework provides methods to quantitatively compare visual explanations using infidelity and sensitivity metrics. This framework can be used by data scientists to perform post-hoc interpretations and explanations of their models, develop more explainable tools and present the findings to clinicians to increase their faith in such models. The proposed framework was evaluated based on a use case scenario of vessel segmentation models trained on Time-of-fight (TOF) Magnetic Resonance Angiogram (MRA) images of the human brain. Quantitative and qualitative results of a comparative study of different models and interpretability methods are presented. Furthermore, this paper provides an extensive overview of several existing interpretability and explainability methods.
Autoregressive (AR) modeling, known for its next-token prediction paradigm, underpins state-of-the-art language and visual generative models. Traditionally, a ``token'' is treated as the smallest prediction unit, often a discrete symbol in language or a quantized patch in vision. However, the optimal token definition for 2D image structures remains an open question. Moreover, AR models suffer from exposure bias, where teacher forcing during training leads to error accumulation at inference. In this paper, we propose xAR, a generalized AR framework that extends the notion of a token to an entity X, which can represent an individual patch token, a cell (a $k\times k$ grouping of neighboring patches), a subsample (a non-local grouping of distant patches), a scale (coarse-to-fine resolution), or even a whole image. Additionally, we reformulate discrete token classification as \textbf{continuous entity regression}, leveraging flow-matching methods at each AR step. This approach conditions training on noisy entities instead of ground truth tokens, leading to Noisy Context Learning, which effectively alleviates exposure bias. As a result, xAR offers two key advantages: (1) it enables flexible prediction units that capture different contextual granularity and spatial structures, and (2) it mitigates exposure bias by avoiding reliance on teacher forcing. On ImageNet-256 generation benchmark, our base model, xAR-B (172M), outperforms DiT-XL/SiT-XL (675M) while achieving 20$\times$ faster inference. Meanwhile, xAR-H sets a new state-of-the-art with an FID of 1.24, running 2.2$\times$ faster than the previous best-performing model without relying on vision foundation modules (\eg, DINOv2) or advanced guidance interval sampling.
Machine learning-based imaging diagnostics has recently reached or even superseded the level of clinical experts in several clinical domains. However, classification decisions of a trained machine learning system are typically non-transparent, a major hindrance for clinical integration, error tracking or knowledge discovery. In this study, we present a transparent deep learning framework relying on convolutional neural networks (CNNs) and layer-wise relevance propagation (LRP) for diagnosing multiple sclerosis (MS). MS is commonly diagnosed utilizing a combination of clinical presentation and conventional magnetic resonance imaging (MRI), specifically the occurrence and presentation of white matter lesions in T2-weighted images. We hypothesized that using LRP in a naive predictive model would enable us to uncover relevant image features that a trained CNN uses for decision-making. Since imaging markers in MS are well-established this would enable us to validate the respective CNN model. First, we pre-trained a CNN on MRI data from the Alzheimer's Disease Neuroimaging Initiative (n = 921), afterwards specializing the CNN to discriminate between MS patients and healthy controls (n = 147). Using LRP, we then produced a heatmap for each subject in the holdout set depicting the voxel-wise relevance for a particular classification decision. The resulting CNN model resulted in a balanced accuracy of 87.04% and an area under the curve of 96.08% in a receiver operating characteristic curve. The subsequent LRP visualization revealed that the CNN model focuses indeed on individual lesions, but also incorporates additional information such as lesion location, non-lesional white matter or gray matter areas such as the thalamus, which are established conventional and advanced MRI markers in MS. We conclude that LRP and the proposed framework have the capability to make diagnostic decisions of...
Generative adversarial networks (GANs) learn a deep generative model that is able to synthesise novel, high-dimensional data samples. New data samples are synthesised by passing latent samples, drawn from a chosen prior distribution, through the generative model. Once trained, the latent space exhibits interesting properties, that may be useful for down stream tasks such as classification or retrieval. Unfortunately, GANs do not offer an "inverse model", a mapping from data space back to latent space, making it difficult to infer a latent representation for a given data sample. In this paper, we introduce a technique, inversion, to project data samples, specifically images, to the latent space using a pre-trained GAN. Using our proposed inversion technique, we are able to identify which attributes of a dataset a trained GAN is able to model and quantify GAN performance, based on a reconstruction loss. We demonstrate how our proposed inversion technique may be used to quantitatively compare performance of various GAN models trained on three image datasets. We provide code for all of our experiments, https://github.com/ToniCreswell/InvertingGAN.
Using deep learning for different machine learning tasks such as image classification and word embedding has recently gained many attentions. Its appealing performance reported across specific Natural Language Processing (NLP) tasks in comparison with other approaches is the reason for its popularity. Word embedding is the task of mapping words or phrases to a low dimensional numerical vector. In this paper, we use deep learning to embed Wikipedia Concepts and Entities. The English version of Wikipedia contains more than five million pages, which suggest its capability to cover many English Entities, Phrases, and Concepts. Each Wikipedia page is considered as a concept. Some concepts correspond to entities, such as a person's name, an organization or a place. Contrary to word embedding, Wikipedia Concepts Embedding is not ambiguous, so there are different vectors for concepts with similar surface form but different mentions. We proposed several approaches and evaluated their performance based on Concept Analogy and Concept Similarity tasks. The results show that proposed approaches have the performance comparable and in some cases even higher than the state-of-the-art methods.
Images of handwritten digits are different from natural images as the orientation of a digit, as well as similarity of features of different digits, makes confusion. On the other hand, deep convolutional neural networks are achieving huge success in computer vision problems, especially in image classification. BDNet is a densely connected deep convolutional neural network model used to classify (recognize) Bengali handwritten numeral digits. It is end-to-end trained using ISI Bengali handwritten numeral dataset. During training, untraditional data preprocessing and augmentation techniques are used so that the trained model works on a different dataset. The model has achieved the test accuracy of 99.775%(baseline was 99.40%) on the test dataset of ISI Bengali handwritten numerals. So, the BDNet model gives 62.5% error reduction compared to previous state-of-the-art models. Here we have also created a dataset of 1000 images of Bengali handwritten numerals to test the trained model, and it giving promising results. Codes, trained model and our own dataset are available at: {https://github.com/Sufianlab/BDNet}.
Feature representation plays a crucial role in visual correspondence, and recent methods for image matching resort to deeply stacked convolutional layers. These models, however, are both monolithic and static in the sense that they typically use a specific level of features, e.g., the output of the last layer, and adhere to it regardless of the images to match. In this work, we introduce a novel approach to visual correspondence that dynamically composes effective features by leveraging relevant layers conditioned on the images to match. Inspired by both multi-layer feature composition in object detection and adaptive inference architectures in classification, the proposed method, dubbed Dynamic Hyperpixel Flow, learns to compose hypercolumn features on the fly by selecting a small number of relevant layers from a deep convolutional neural network. We demonstrate the effectiveness on the task of semantic correspondence, i.e., establishing correspondences between images depicting different instances of the same object or scene category. Experiments on standard benchmarks show that the proposed method greatly improves matching performance over the state of the art in an adaptive and efficient manner.
This paper introduces data augmentation for point clouds by interpolation between examples. Data augmentation by interpolation has shown to be a simple and effective approach in the image domain. Such a mixup is however not directly transferable to point clouds, as we do not have a one-to-one correspondence between the points of two different objects. In this paper, we define data augmentation between point clouds as a shortest path linear interpolation. To that end, we introduce PointMixup, an interpolation method that generates new examples through an optimal assignment of the path function between two point clouds. We prove that our PointMixup finds the shortest path between two point clouds and that the interpolation is assignment invariant and linear. With the definition of interpolation, PointMixup allows to introduce strong interpolation-based regularizers such as mixup and manifold mixup to the point cloud domain. Experimentally, we show the potential of PointMixup for point cloud classification, especially when examples are scarce, as well as increased robustness to noise and geometric transformations to points. The code for PointMixup and the experimental details are publicly available.
Detecting latent structure within a dataset is a crucial step in performing analysis of a dataset. However, existing state-of-the-art techniques for subclass discovery are limited: either they are limited to detecting very small numbers of outliers or they lack the statistical power to deal with complex data such as image or audio. This paper proposes a solution to this subclass discovery problem: by leveraging instance explanation methods, an existing classifier can be extended to detect latent classes via differences in the classifier's internal decisions about each instance. This works not only with simple classification techniques but also with deep neural networks, allowing for a powerful and flexible approach to detecting latent structure within datasets. Effectively, this represents a projection of the dataset into the classifier's "explanation space," and preliminary results show that this technique outperforms the baseline for the detection of latent classes even with limited processing. This paper also contains a pipeline for analyzing classifiers automatically, and a web application for interactively exploring the results from this technique.
We propose contextual convolution (CoConv) for visual recognition. CoConv is a direct replacement of the standard convolution, which is the core component of convolutional neural networks. CoConv is implicitly equipped with the capability of incorporating contextual information while maintaining a similar number of parameters and computational cost compared to the standard convolution. CoConv is inspired by neuroscience studies indicating that (i) neurons, even from the primary visual cortex (V1 area), are involved in detection of contextual cues and that (ii) the activity of a visual neuron can be influenced by the stimuli placed entirely outside of its theoretical receptive field. On the one hand, we integrate CoConv in the widely-used residual networks and show improved recognition performance over baselines on the core tasks and benchmarks for visual recognition, namely image classification on the ImageNet data set and object detection on the MS COCO data set. On the other hand, we introduce CoConv in the generator of a state-of-the-art Generative Adversarial Network, showing improved generative results on CIFAR-10 and CelebA. Our code is available at https://github.com/iduta/coconv.
While deep learning has been very beneficial in data-rich settings, tasks with smaller training set often resort to pre-training or multitask learning to leverage data from other tasks. In this case, careful consideration is needed to select tasks and model parameterizations such that updates from the auxiliary tasks actually help the primary task. We seek to alleviate this burden by formulating a model-agnostic framework that performs fine-grained manipulation of the auxiliary task gradients. We propose to decompose auxiliary updates into directions which help, damage or leave the primary task loss unchanged. This allows weighting the update directions differently depending on their impact on the problem of interest. We present a novel and efficient algorithm for that purpose and show its advantage in practice. Our method leverages efficient automatic differentiation procedures and randomized singular value decomposition for scalability. We show that our framework is generic and encompasses some prior work as particular cases. Our approach consistently outperforms strong and widely used baselines when leveraging out-of-distribution data for Text and Image classification tasks.
Out-of-distribution (OOD) detection has received much attention lately due to its importance in the safe deployment of neural networks. One of the key challenges is that models lack supervision signals from unknown data, and as a result, can produce overconfident predictions on OOD data. Previous approaches rely on real outlier datasets for model regularization, which can be costly and sometimes infeasible to obtain in practice. In this paper, we present VOS, a novel framework for OOD detection by adaptively synthesizing virtual outliers that can meaningfully regularize the model's decision boundary during training. Specifically, VOS samples virtual outliers from the low-likelihood region of the class-conditional distribution estimated in the feature space. Alongside, we introduce a novel unknown-aware training objective, which contrastively shapes the uncertainty space between the ID data and synthesized outlier data. VOS achieves competitive performance on both object detection and image classification models, reducing the FPR95 by up to 9.36% compared to the previous best method on object detectors. Code is available at https://github.com/deeplearning-wisc/vos.
Graph Convolutional Networks (GCNs) achieve an impressive performance due to the remarkable representation ability in learning the graph information. However, GCNs, when implemented on a deep network, require expensive computation power, making them difficult to be deployed on battery-powered devices. In contrast, Spiking Neural Networks (SNNs), which perform a bio-fidelity inference process, offer an energy-efficient neural architecture. In this work, we propose SpikingGCN, an end-to-end framework that aims to integrate the embedding of GCNs with the biofidelity characteristics of SNNs. The original graph data are encoded into spike trains based on the incorporation of graph convolution. We further model biological information processing by utilizing a fully connected layer combined with neuron nodes. In a wide range of scenarios (e.g. citation networks, image graph classification, and recommender systems), our experimental results show that the proposed method could gain competitive performance against state-of-the-art approaches. Furthermore, we show that SpikingGCN on a neuromorphic chip can bring a clear advantage of energy efficiency into graph data analysis, which demonstrates its great potential to construct environment-friendly machine learning models.
Whilst computer vision models built using self-supervised approaches are now commonplace, some important questions remain. Do self-supervised models learn highly redundant channel features? What if a self-supervised network could dynamically select the important channels and get rid of the unnecessary ones? Currently, convnets pre-trained with self-supervision have obtained comparable performance on downstream tasks in comparison to their supervised counterparts in computer vision. However, there are drawbacks to self-supervised models including their large numbers of parameters, computationally expensive training strategies and a clear need for faster inference on downstream tasks. In this work, our goal is to address the latter by studying how a standard channel selection method developed for supervised learning can be applied to networks trained with self-supervision. We validate our findings on a range of target budgets $t_{d}$ for channel computation on image classification task across different datasets, specifically CIFAR-10, CIFAR-100, and ImageNet-100, obtaining comparable performance to that of the original network when selecting all channels but at a significant reduction in computation reported in terms of FLOPs.
Large-scale well-annotated datasets are of great importance for training an effective object detector. However, obtaining accurate bounding box annotations is laborious and demanding. Unfortunately, the resultant noisy bounding boxes could cause corrupt supervision signals and thus diminish detection performance. Motivated by the observation that the real ground-truth is usually situated in the aggregation region of the proposals assigned to a noisy ground-truth, we propose DIStribution-aware CalibratiOn (DISCO) to model the spatial distribution of proposals for calibrating supervision signals. In DISCO, spatial distribution modeling is performed to statistically extract the potential locations of objects. Based on the modeled distribution, three distribution-aware techniques, i.e., distribution-aware proposal augmentation (DA-Aug), distribution-aware box refinement (DA-Ref), and distribution-aware confidence estimation (DA-Est), are developed to improve classification, localization, and interpretability, respectively. Extensive experiments on large-scale noisy image datasets (i.e., Pascal VOC and MS-COCO) demonstrate that DISCO can achieve state-of-the-art detection performance, especially at high noise levels. Code is available at https://github.com/Correr-Zhou/DISCO.
Continual learning is a challenging problem in which models need to be trained on non-stationary data across sequential tasks for class-incremental learning. While previous methods have focused on using either regularization or rehearsal-based frameworks to alleviate catastrophic forgetting in image classification, they are limited to a single modality and cannot learn compact class-aware cross-modal representations for continual audio-visual learning. To address this gap, we propose a novel class-incremental grouping network (CIGN) that can learn category-wise semantic features to achieve continual audio-visual learning. Our CIGN leverages learnable audio-visual class tokens and audio-visual grouping to continually aggregate class-aware features. Additionally, it utilizes class tokens distillation and continual grouping to prevent forgetting parameters learned from previous tasks, thereby improving the model's ability to capture discriminative audio-visual categories. We conduct extensive experiments on VGGSound-Instruments, VGGSound-100, and VGG-Sound Sources benchmarks. Our experimental results demonstrate that the CIGN achieves state-of-the-art audio-visual class-incremental learning performance. Code is available at https://github.com/stoneMo/CIGN.