Search

Vision Transformers (ViTs) have shown success across a variety of tasks due to their ability to capture global image representations. Recent studies have identified the existence of high-norm tokens in ViTs, which can interfere with unsupervised object discovery. To address this, the use of "registers" which are additional tokens that isolate high norm patch tokens while capturing global image-level information has been proposed. While registers have been studied extensively for object discovery, their generalization properties particularly in out-of-distribution (OOD) scenarios, remains underexplored. In this paper, we examine the utility of register token embeddings in providing additional features for improving generalization and anomaly rejection. To that end, we propose a simple method that combines the special CLS token embedding commonly employed in ViTs with the average-pooled register embeddings to create feature representations which are subsequently used for training a downstream classifier. We find that this enhances OOD generalization and anomaly rejection, while maintaining in-distribution (ID) performance. Extensive experiments across multiple ViT backbones trained with and without registers reveal consistent improvements of 2-4\% in top-1 OOD accuracy and a 2-3\% reduction in false positive rates for anomaly detection. Importantly, these gains are achieved without additional computational overhead., Comment: Accepted at ICASSP 2025

Reliably detecting when a deployed machine learning model is likely to fail on a given input is crucial for ensuring safe operation. In this work, we propose DECIDER (Debiasing Classifiers to Identify Errors Reliably), a novel approach that leverages priors from large language models (LLMs) and vision-language models (VLMs) to detect failures in image classification models. DECIDER utilizes LLMs to specify task-relevant core attributes and constructs a ``debiased'' version of the classifier by aligning its visual features to these core attributes using a VLM, and detects potential failure by measuring disagreement between the original and debiased models. In addition to proactively identifying samples on which the model would fail, DECIDER also provides human-interpretable explanations for failure through a novel attribute-ablation strategy. Through extensive experiments across diverse benchmarks spanning subpopulation shifts (spurious correlations, class imbalance) and covariate shifts (synthetic corruptions, domain shifts), DECIDER consistently achieves state-of-the-art failure detection performance, significantly outperforming baselines in terms of the overall Matthews correlation coefficient as well as failure and success recall. Our codes can be accessed at~\url{https://github.com/kowshikthopalli/DECIDER/}, Comment: Accepted at ECCV (European Conference on Computer Vision) 2024

Anchoring is a recent, architecture-agnostic principle for training deep neural networks that has been shown to significantly improve uncertainty estimation, calibration, and extrapolation capabilities. In this paper, we systematically explore anchoring as a general protocol for training vision models, providing fundamental insights into its training and inference processes and their implications for generalization and safety. Despite its promise, we identify a critical problem in anchored training that can lead to an increased risk of learning undesirable shortcuts, thereby limiting its generalization capabilities. To address this, we introduce a new anchored training protocol that employs a simple regularizer to mitigate this issue and significantly enhances generalization. We empirically evaluate our proposed approach across datasets and architectures of varying scales and complexities, demonstrating substantial performance gains in generalization and safety metrics compared to the standard training protocol.

Safe deployment of AI models requires proactive detection of failures to prevent costly errors. To this end, we study the important problem of detecting failures in deep regression models. Existing approaches rely on epistemic uncertainty estimates or inconsistency w.r.t the training data to identify failure. Interestingly, we find that while uncertainties are necessary they are insufficient to accurately characterize failure in practice. Hence, we introduce PAGER (Principled Analysis of Generalization Errors in Regressors), a framework to systematically detect and characterize failures in deep regressors. Built upon the principle of anchored training in deep models, PAGER unifies both epistemic uncertainty and complementary manifold non-conformity scores to accurately organize samples into different risk regimes., Comment: Published at ICML 2024

We propose an accurate and fast classification network for classification of brain tumors in MRI images that outperforms all lightweight methods investigated in terms of accuracy. We test our model on a challenging 2D T1-weighted CE-MRI dataset containing three types of brain tumors: Meningioma, Glioma and Pituitary. We introduce an l2-normalized spatial attention mechanism that acts as a regularizer against overfitting during training. We compare our results against the state-of-the-art on this dataset and show that by integrating l2-normalized spatial attention into a baseline network we achieve a performance gain of 1.79 percentage points. Even better accuracy can be attained by combining our model in an ensemble with the pretrained VGG16 at the expense of execution speed. Our code is publicly available at https://github.com/juliadietlmeier/MRI_image_classification, Comment: Accepted to be published in: IEEE International Conference on Image Processing (ICIP), Kuala Lumpur October 8-11, 2023

We are interested in estimating the uncertainties of deep neural networks, which play an important role in many scientific and engineering problems. In this paper, we present a striking new finding that an ensemble of neural networks with the same weight initialization, trained on datasets that are shifted by a constant bias gives rise to slightly inconsistent trained models, where the differences in predictions are a strong indicator of epistemic uncertainties. Using the neural tangent kernel (NTK), we demonstrate that this phenomena occurs in part because the NTK is not shift-invariant. Since this is achieved via a trivial input transformation, we show that this behavior can therefore be approximated by training a single neural network -- using a technique that we call $\Delta-$UQ -- that estimates uncertainty around prediction by marginalizing out the effect of the biases during inference. We show that $\Delta-$UQ's uncertainty estimates are superior to many of the current methods on a variety of benchmarks -- outlier rejection, calibration under distribution shift, and sequential design optimization of black box functions. Code for $\Delta-$UQ can be accessed at https://github.com/LLNL/DeltaUQ, Comment: Spotlight at NeurIPS 2022

We focus on the problem of producing well-calibrated out-of-distribution (OOD) detectors, in order to enable safe deployment of medical image classifiers. Motivated by the difficulty of curating suitable calibration datasets, synthetic augmentations have become highly prevalent for inlier/outlier specification. While there have been rapid advances in data augmentation techniques, this paper makes a striking finding that the space in which the inliers and outliers are synthesized, in addition to the type of augmentation, plays a critical role in calibrating OOD detectors. Using the popular energy-based OOD detection framework, we find that the optimal protocol is to synthesize latent-space inliers along with diverse pixel-space outliers. Based on empirical studies with multiple medical imaging benchmarks, we demonstrate that our approach consistently leads to superior OOD detection ($15\% - 35\%$ in AUROC) over the state-of-the-art in a variety of open-set recognition settings.

Explanation techniques that synthesize small, interpretable changes to a given image while producing desired changes in the model prediction have become popular for introspecting black-box models. Commonly referred to as counterfactuals, the synthesized explanations are required to contain discernible changes (for easy interpretability) while also being realistic (consistency to the data manifold). In this paper, we focus on the case where we have access only to the trained deep classifier and not the actual training data. While the problem of inverting deep models to synthesize images from the training distribution has been explored, our goal is to develop a deep inversion approach to generate counterfactual explanations for a given query image. Despite their effectiveness in conditional image synthesis, we show that existing deep inversion methods are insufficient for producing meaningful counterfactuals. We propose DISC (Deep Inversion for Synthesizing Counterfactuals) that improves upon deep inversion by utilizing (a) stronger image priors, (b) incorporating a novel manifold consistency objective and (c) adopting a progressive optimization strategy. We find that, in addition to producing visually meaningful explanations, the counterfactuals from DISC are effective at learning classifier decision boundaries and are robust to unknown test-time corruptions., Comment: Neurips 2021

Unsupervised deep learning methods for solving audio restoration problems extensively rely on carefully tailored neural architectures that carry strong inductive biases for defining priors in the time or spectral domain. In this context, lot of recent success has been achieved with sophisticated convolutional network constructions that recover audio signals in the spectral domain. However, in practice, audio priors require careful engineering of the convolutional kernels to be effective at solving ill-posed restoration tasks, while also being easy to train. To this end, in this paper, we propose a new U-Net based prior that does not impact either the network complexity or convergence behavior of existing convolutional architectures, yet leads to significantly improved restoration. In particular, we advocate the use of carefully designed dilation schedules and dense connections in the U-Net architecture to obtain powerful audio priors. Using empirical studies on standard benchmarks and a variety of ill-posed restoration tasks, such as audio denoising, in-painting and source separation, we demonstrate that our proposed approach consistently outperforms widely adopted audio prior architectures.

With increased interest in adopting AI methods for clinical diagnosis, a vital step towards safe deployment of such tools is to ensure that the models not only produce accurate predictions but also do not generalize to data regimes where the training data provide no meaningful evidence. Existing approaches for ensuring the distribution of model predictions to be similar to that of the true distribution rely on explicit uncertainty estimators that are inherently hard to calibrate. In this paper, we propose to train a loss estimator alongside the predictive model, using a contrastive training objective, to directly estimate the prediction uncertainties. Interestingly, we find that, in addition to producing well-calibrated uncertainties, this approach improves the generalization behavior of the predictor. Using a dermatology use-case, we show the impact of loss estimators on model generalization, in terms of both its fidelity on in-distribution data and its ability to detect out of distribution samples or new classes unseen during training.

Through the use of carefully tailored convolutional neural network architectures, a deep image prior (DIP) can be used to obtain pre-images from latent representation encodings. Though DIP inversion has been known to be superior to conventional regularized inversion strategies such as total variation, such an over-parameterized generator is able to effectively reconstruct even images that are not in the original data distribution. This limitation makes it challenging to utilize such priors for tasks such as counterfactual reasoning, wherein the goal is to generate small, interpretable changes to an image that systematically leads to changes in the model prediction. To this end, we propose a novel regularization strategy based on an auxiliary loss estimator jointly trained with the predictor, which efficiently guides the prior to recover natural pre-images. Our empirical studies with a real-world ISIC skin lesion detection problem clearly evidence the effectiveness of the proposed approach in synthesizing meaningful counterfactuals. In comparison, we find that the standard DIP inversion often proposes visually imperceptible perturbations to irrelevant parts of the image, thus providing no additional insights into the model behavior.

With increasing reliance on the outcomes of black-box models in critical applications, post-hoc explainability tools that do not require access to the model internals are often used to enable humans understand and trust these models. In particular, we focus on the class of methods that can reveal the influence of input features on the predicted outputs. Despite their wide-spread adoption, existing methods are known to suffer from one or more of the following challenges: computational complexities, large uncertainties and most importantly, inability to handle real-world domain shifts. In this paper, we propose PRoFILE, a novel feature importance estimation method that addresses all these challenges. Through the use of a loss estimator jointly trained with the predictive model and a causal objective, PRoFILE can accurately estimate the feature importance scores even under complex distribution shifts, without any additional re-training. To this end, we also develop learning strategies for training the loss estimator, namely contrastive and dropout calibration, and find that it can effectively detect distribution shifts. Using empirical studies on several benchmark image and non-image data, we show significant improvements over state-of-the-art approaches, both in terms of fidelity and robustness.

State-of-the-art under-determined audio source separation systems rely on supervised end-end training of carefully tailored neural network architectures operating either in the time or the spectral domain. However, these methods are severely challenged in terms of requiring access to expensive source level labeled data and being specific to a given set of sources and the mixing process, which demands complete re-training when those assumptions change. This strongly emphasizes the need for unsupervised methods that can leverage the recent advances in data-driven modeling, and compensate for the lack of labeled data through meaningful priors. To this end, we propose a novel approach for audio source separation based on generative priors trained on individual sources. Through the use of projected gradient descent optimization, our approach simultaneously searches in the source-specific latent spaces to effectively recover the constituent sources. Though the generative priors can be defined in the time domain directly, e.g. WaveGAN, we find that using spectral domain loss functions for our optimization leads to good-quality source estimates. Our empirical studies on standard spoken digit and instrument datasets clearly demonstrate the effectiveness of our approach over classical as well as state-of-the-art unsupervised baselines., Comment: 5 pages, 2 figures

Modern audio source separation techniques rely on optimizing sequence model architectures such as, 1D-CNNs, on mixture recordings to generalize well to unseen mixtures. Specifically, recent focus is on time-domain based architectures such as Wave-U-Net which exploit temporal context by extracting multi-scale features. However, the optimality of the feature extraction process in these architectures has not been well investigated. In this paper, we examine and recommend critical architectural changes that forge an optimal multi-scale feature extraction process. To this end, we replace regular $1-$D convolutions with adaptive dilated convolutions that have innate capability of capturing increased context by using large temporal receptive fields. We also investigate the impact of dense connections on the extraction process that encourage feature reuse and better gradient flow. The dense connections between the downsampling and upsampling paths of a U-Net architecture capture multi-resolution information leading to improved temporal modelling. We evaluate the proposed approaches on the MUSDB test dataset. In addition to providing an improved performance over the state-of-the-art, we also provide insights on the impact of different architectural choices on complex data-driven solutions for source separation.

State-of-the-art speaker diarization systems utilize knowledge from external data, in the form of a pre-trained distance metric, to effectively determine relative speaker identities to unseen data. However, much of recent focus has been on choosing the appropriate feature extractor, ranging from pre-trained $i-$vectors to representations learned via different sequence modeling architectures (e.g. 1D-CNNs, LSTMs, attention models), while adopting off-the-shelf metric learning solutions. In this paper, we argue that, regardless of the feature extractor, it is crucial to carefully design a metric learning pipeline, namely the loss function, the sampling strategy and the discrimnative margin parameter, for building robust diarization systems. Furthermore, we propose to adopt a fine-grained validation process to obtain a comprehensive evaluation of the generalization power of metric learning pipelines. To this end, we measure diarization performance across different language speakers, and variations in the number of speakers in a recording. Using empirical studies, we provide interesting insights into the effectiveness of different design choices and make recommendations.