COOkeD: Ensemble-based OOD detection in the era of zero-shot CLIP

Out-of-distribution (OOD) detection is an important building block in trustworthy image recognition systems as unknown classes may arise at test-time. OOD detection methods typically revolve around a single classifier, leading to a split in the research field between the classical supervised setting (e.g. ResNet18 classifier trained on CIFAR100) vs. the zero-shot setting (class names fed as prompts to CLIP). In both cases, an overarching challenge is that the OOD detection performance is implicitly constrained by the classifier's capabilities on in-distribution (ID) data. In this work, we show that given a little open-mindedness from both ends, remarkable OOD detection can be achieved by instead creating a heterogeneous ensemble - COOkeD combines the predictions of a closed-world classifier trained end-to-end on a specific dataset, a zero-shot CLIP classifier, and a linear probe classifier trained on CLIP image features. While bulky at first sight, this approach is modular, post-hoc and leverages the availability of pre-trained VLMs, thus introduces little overhead compared to training a single standard classifier. We evaluate COOkeD on popular CIFAR100 and ImageNet benchmarks, but also consider more challenging, realistic settings ranging from training-time label noise, to test-time covariate shift, to zero-shot shift which has been previously overlooked. Despite its simplicity, COOkeD achieves state-of-the-art performance and greater robustness compared to both classical and CLIP-based OOD detection methods. Code is available at https://github.com/glhr/COOkeD

Explainability for Vision Foundation Models: A Survey

As artificial intelligence systems become increasingly integrated into daily life, the field of explainability has gained significant attention. This trend is particularly driven by the complexity of modern AI models and their decision-making processes. The advent of foundation models, characterized by their extensive generalization capabilities and emergent uses, has further complicated this landscape. Foundation models occupy an ambiguous position in the explainability domain: their complexity makes them inherently challenging to interpret, yet they are increasingly leveraged as tools to construct explainable models. In this survey, we explore the intersection of foundation models and eXplainable AI (XAI) in the vision domain. We begin by compiling a comprehensive corpus of papers that bridge these fields. Next, we categorize these works based on their architectural characteristics. We then discuss the challenges faced by current research in integrating XAI within foundation models. Furthermore, we review common evaluation methodologies for these combined approaches. Finally, we present key observations and insights from our survey, offering directions for future research in this rapidly evolving field.

Towards Understanding and Quantifying Uncertainty for Text-to-Image Generation

Uncertainty quantification in text-to-image (T2I) generative models is crucial for understanding model behavior and improving output reliability. In this paper, we are the first to quantify and evaluate the uncertainty of T2I models with respect to the prompt. Alongside adapting existing approaches designed to measure uncertainty in the image space, we also introduce Prompt-based UNCertainty Estimation for T2I models (PUNC), a novel method leveraging Large Vision-Language Models (LVLMs) to better address uncertainties arising from the semantics of the prompt and generated images. PUNC utilizes a LVLM to caption a generated image, and then compares the caption with the original prompt in the more semantically meaningful text space. PUNC also enables the disentanglement of both aleatoric and epistemic uncertainties via precision and recall, which image-space approaches are unable to do. Extensive experiments demonstrate that PUNC outperforms state-of-the-art uncertainty estimation techniques across various settings. Uncertainty quantification in text-to-image generation models can be used on various applications including bias detection, copyright protection, and OOD detection. We also introduce a comprehensive dataset of text prompts and generation pairs to foster further research in uncertainty quantification for generative models. Our findings illustrate that PUNC not only achieves competitive performance but also enables novel applications in evaluating and improving the trustworthiness of text-to-image models.

Collaborative Instance Navigation: Leveraging Agent Self-Dialogue to Minimize User Input

Existing embodied instance goal navigation tasks, driven by natural language, assume human users to provide complete and nuanced instance descriptions prior to the navigation, which can be impractical in the real world as human instructions might be brief and ambiguous. To bridge this gap, we propose a new task, Collaborative Instance Navigation (CoIN), with dynamic agent-human interaction during navigation to actively resolve uncertainties about the target instance in natural, template-free, open-ended dialogues. To address CoIN, we propose a novel method, Agent-user Interaction with UncerTainty Awareness (AIUTA), leveraging the perception capability of Vision Language Models (VLMs) and the capability of Large Language Models (LLMs). First, upon object detection, a Self-Questioner model initiates a self-dialogue to obtain a complete and accurate observation description, while a novel uncertainty estimation technique mitigates inaccurate VLM perception. Then, an Interaction Trigger module determines whether to ask a question to the user, continue or halt navigation, minimizing user input. For evaluation, we introduce CoIN-Bench, a benchmark supporting both real and simulated humans. AIUTA achieves competitive performance in instance navigation against state-of-the-art methods, demonstrating great flexibility in handling user inputs.

Double Descent Meets Out-of-Distribution Detection: Theoretical Insights and Empirical Analysis on the role of model complexity

While overparameterization is known to benefit generalization, its impact on Out-Of-Distribution (OOD) detection is less understood. This paper investigates the influence of model complexity in OOD detection. We propose an expected OOD risk metric to evaluate classifiers confidence on both training and OOD samples. Leveraging Random Matrix Theory, we derive bounds for the expected OOD risk of binary least-squares classifiers applied to Gaussian data. We show that the OOD risk depicts an infinite peak, when the number of parameters is equal to the number of samples, which we associate with the double descent phenomenon. Our experimental study on different OOD detection methods across multiple neural architectures extends our theoretical insights and highlights a double descent curve. Our observations suggest that overparameterization does not necessarily lead to better OOD detection. Using the Neural Collapse framework, we provide insights to better understand this behavior. To facilitate reproducibility, our code will be made publicly available upon publication.

Benchmarking XAI Explanations with Human-Aligned Evaluations

In this paper, we introduce PASTA (Perceptual Assessment System for explanaTion of Artificial intelligence), a novel framework for a human-centric evaluation of XAI techniques in computer vision. Our first key contribution is a human evaluation of XAI explanations on four diverse datasets (COCO, Pascal Parts, Cats Dogs Cars, and MonumAI) which constitutes the first large-scale benchmark dataset for XAI, with annotations at both the image and concept levels. This dataset allows for robust evaluation and comparison across various XAI methods. Our second major contribution is a data-based metric for assessing the interpretability of explanations. It mimics human preferences, based on a database of human evaluations of explanations in the PASTA-dataset. With its dataset and metric, the PASTA framework provides consistent and reliable comparisons between XAI techniques, in a way that is scalable but still aligned with human evaluations. Additionally, our benchmark allows for comparisons between explanations across different modalities, an aspect previously unaddressed. Our findings indicate that humans tend to prefer saliency maps over other explanation types. Moreover, we provide evidence that human assessments show a low correlation with existing XAI metrics that are numerically simulated by probing the model.

Organizing Unstructured Image Collections using Natural Language

Organizing unstructured visual data into semantic clusters is a key challenge in computer vision. Traditional deep clustering (DC) approaches focus on a single partition of data, while multiple clustering (MC) methods address this limitation by uncovering distinct clustering solutions. The rise of large language models (LLMs) and multimodal LLMs (MLLMs) has enhanced MC by allowing users to define clustering criteria in natural language. However, manually specifying criteria for large datasets is impractical. In this work, we introduce the task Semantic Multiple Clustering (SMC) that aims to automatically discover clustering criteria from large image collections, uncovering interpretable substructures without requiring human input. Our framework, Text Driven Semantic Multiple Clustering (TeDeSC), uses text as a proxy to concurrently reason over large image collections, discover partitioning criteria, expressed in natural language, and reveal semantic substructures. To evaluate TeDeSC, we introduce the COCO-4c and Food-4c benchmarks, each containing four grouping criteria and ground-truth annotations. We apply TeDeSC to various applications, such as discovering biases and analyzing social media image popularity, demonstrating its utility as a tool for automatically organizing image collections and revealing novel insights.

Frustratingly Easy Test-Time Adaptation of Vision-Language Models

Vision-Language Models seamlessly discriminate among arbitrary semantic categories, yet they still suffer from poor generalization when presented with challenging examples. For this reason, Episodic Test-Time Adaptation (TTA) strategies have recently emerged as powerful techniques to adapt VLMs in the presence of a single unlabeled image. The recent literature on TTA is dominated by the paradigm of prompt tuning by Marginal Entropy Minimization, which, relying on online backpropagation, inevitably slows down inference while increasing memory. In this work, we theoretically investigate the properties of this approach and unveil that a surprisingly strong TTA method lies dormant and hidden within it. We term this approach ZERO (TTA with "zero" temperature), whose design is both incredibly effective and frustratingly simple: augment N times, predict, retain the most confident predictions, and marginalize after setting the Softmax temperature to zero. Remarkably, ZERO requires a single batched forward pass through the vision encoder only and no backward passes. We thoroughly evaluate our approach following the experimental protocol established in the literature and show that ZERO largely surpasses or compares favorably w.r.t. the state-of-the-art while being almost 10x faster and 13x more memory-friendly than standard Test-Time Prompt Tuning. Thanks to its simplicity and comparatively negligible computation, ZERO can serve as a strong baseline for future work in this field. The code is available at https://github.com/FarinaMatteo/zero.

Hierarchical Light Transformer Ensembles for Multimodal Trajectory Forecasting

Accurate trajectory forecasting is crucial for the performance of various systems, such as advanced driver-assistance systems and self-driving vehicles. These forecasts allow to anticipate events leading to collisions and, therefore, to mitigate them. Deep Neural Networks have excelled in motion forecasting, but issues like overconfidence and uncertainty quantification persist. Deep Ensembles address these concerns, yet applying them to multimodal distributions remains challenging. In this paper, we propose a novel approach named Hierarchical Light Transformer Ensembles (HLT-Ens), aimed at efficiently training an ensemble of Transformer architectures using a novel hierarchical loss function. HLT-Ens leverages grouped fully connected layers, inspired by grouped convolution techniques, to capture multimodal distributions, effectively. Through extensive experimentation, we demonstrate that HLT-Ens achieves state-of-the-art performance levels, offering a promising avenue for improving trajectory forecasting techniques.

Understanding Why Label Smoothing Degrades Selective Classification and How to Fix It

Label smoothing (LS) is a popular regularisation method for training deep neural network classifiers due to its effectiveness in improving test accuracy and its simplicity in implementation. "Hard" one-hot labels are "smoothed" by uniformly distributing probability mass to other classes, reducing overfitting. In this work, we reveal that LS negatively affects selective classification (SC) - where the aim is to reject misclassifications using a model's predictive uncertainty. We first demonstrate empirically across a range of tasks and architectures that LS leads to a consistent degradation in SC. We then explain this by analysing logit-level gradients, showing that LS exacerbates overconfidence and underconfidence by regularising the max logit more when the probability of error is low, and less when the probability of error is high. This elucidates previously reported experimental results where strong classifiers underperform in SC. We then demonstrate the empirical effectiveness of logit normalisation for recovering lost SC performance caused by LS. Furthermore, based on our gradient analysis, we explain why such normalisation is effective. We will release our code shortly.