Images and structured tables are essential parts of real-world databases. Though tabular-image representation learning is promising to create new insights, it remains a challenging task, as tabular data is typically heterogeneous and incomplete, presenting significant modality disparities with images. Earlier works have mainly focused on simple modality fusion strategies in complete data scenarios, without considering the missing data issue, and thus are limited in practice. In this paper, we propose TIP, a novel tabular-image pre-training framework for learning multimodal representations robust to incomplete tabular data. Specifically, TIP investigates a novel self-supervised learning (SSL) strategy, including a masked tabular reconstruction task for tackling data missingness, and image-tabular matching and contrastive learning objectives to capture multimodal information. Moreover, TIP proposes a versatile tabular encoder tailored for incomplete, heterogeneous tabular data and a multimodal interaction module for inter-modality representation learning. Experiments are performed on downstream multimodal classification tasks using both natural and medical image datasets. The results show that TIP outperforms state-of-the-art supervised/SSL image/multimodal algorithms in both complete and incomplete data scenarios. Our code will be available at https://github.com/anonymous.
论文检索
输入标题、作者或关键词，从 7,876 篇学术成果中精准定位
DECIDER: Leveraging Foundation Model Priors for Improved Model Failure Detection and Explanation
PDF ↗In this paper, we focus on the problem of detecting samples that can lead to model failure under the classification setting. Failures can stem from various sources, such as spurious correlations between image features and labels, class imbalances in the training data, and covariate shifts between training and test distributions. Existing approaches often rely on classifier prediction scores and do not comprehensively identify all failure scenarios. Instead, we pose failure detection as the problem of identifying the discrepancies between the classifier and its enhanced version. We build such an enhanced model by infusing task-agnostic prior knowledge from a vision-language model (e.g., CLIP) that encodes general-purpose visual and semantic relationships. Unlike conventional training, our enhanced model, named the Prior Induced Model (PIM) learns to map the pre-trained model features to the VLM latent space and aligns the same with a set of pre-specified, fine-grained class-level attributes which are later aggregated to estimate the class prediction. We propose that such a training strategy allows the model to concentrate only on the task specific attributes while making predictions in lieu of the pre-trained model and also enables human-interpretable explanations for failure. We conduct extensive empirical studies on various benchmark datasets and baselines, observing substantial improvements in failure detection.
In this work, we focus on the cross-domain few-shot classification (CDFSC), which is mostly challenged by the low-data problem as well as extreme domain shift between base and novel target classes. Current methods always employ a lightweight backbone and continue to use a linear-probe-like traditional fine-tuning (Trad-FT) paradigm. While for recently emerging large-scale pre-trained model (LPM), which has more parameters with considerable prior knowledge, employing Trad-FT will face significant risks of overfitting and prior knowledge damage. In this paper, we propose semantic-guided robustness tuning (SRT), a novel fine-tuning paradigm including modulus-matching-based image-text mixup (MMIT-Mixup) and robustness-invariance fine-tuning (RI-FT), to address the CDFSC challenge of LPM. Concretely, SRT focuses on achieving robust class-specific representation. It first considers textual information as a robust and domain-invariant conductor, and MMIT-Mixup injects the domain-invariant and class-specific knowledge to obtain domain-invariant prototypes. Then, RI-FT optimizes the distance between features and prototypes to enhance the robustness of visual-encoder. We consider several types of LPMs and conduct extensive experiments, which reveals that SRT is a general solution for LPM’s CDFSC challenge and outperforms the existing methods with a large margin.
Earth Observation (EO) presents a unique opportunity to explore self-supervised multimodal learning, given its access to vast and diverse data captured by various sensors. However, current multimodal EO datasets and models often consider modalities from a single data type, either mono-date images or time series, which limits their expressivity. We introduce OmniSat, a novel architecture that exploits the natural alignment between multiple EO modalities to learn expressive multimodal representations without labels. We augment an existing dataset with new modalities to demonstrate the advantages of combining modalities of different natures. We evaluate OmniSat and various state-of-the-art approaches on two relevant downstream tasks: forestry and land cover classification. Our results show that OmniSat can learn rich representations in an unsupervised manner, leading to performance improvements in the semi- and fully-supervised settings, even when only one modality is available at inference. Our code, weights, and dataset are available at https://github.com/gastruc/OmniSat.
Vision-and-language foundation models have shown impressive hallmarks on zero-shot image classification, where the target classes are represented in text descriptions with no labeled image examples. Recent work spans such powerful image and text correspondence to open-vocabulary segmentation, \ie, predicting pixel and text correspondence without pixel-level supervision on the unseen target classes. Plenty of the previous art casts this task as pixel-to-text classification without the goal of comprehending objects within an image. We believe segmentation is a visual understanding task and advocate decoupling segmentation from visual grounding. To this end, we introduce Lazy Visual Grounding for zero-shot open-vocabulary segmentation. Lazy visual grounding first discovers distinguishable visual units as object masks with iterative graph cuts and then assigns text on the discovered visual objects in a late interaction manner. Our model is training-free yet shows great performance on four public datasets: Pascal VOC, COCO-object, COCO-stuff, and ADE 20K, and especially, demonstrates visually appealing segmentation results, indicating the model capability to comprehend visual objectness. Code and data will be released once accepted.
Few-Shot Class-Incremental Learning (FSCIL) aims to learn new concepts with few training samples while preserving previously acquired knowledge. Although promising performance has been achieved, there remains an underexplored aspect regarding the basic statistical principles underlying FSCIL. Therefore, we thoroughly explore the approximation risk of FSCIL, encompassing both transfer and consistency risks. By tightening the upper bounds of these risks, we derive practical guidelines for designing and training FSCIL models. These guidelines include (1) expanding training datasets for base classes, (2) preventing excessive focus on specific features, (3) optimizing classification margin discrepancy, and (4) ensuring unbiased classification across both base and novel classes. Leveraging these insights, we conduct comprehensive experiments to validate our principles, achieving state-of-the-art performance on three FSCIL benchmark datasets.
Self-supervised learning (SSL) is critical for learning high-quality representations from unlabeled images at scale. Earlier efforts at reducing the compute requirements of SSL have focused on identifying subsets of training data that are sufficient for training. In addition to using a static representative subset, these methods also require small amounts of labeled data for scoring instances. In this work, we design a new family of algorithms that exploits the training dynamics of SSL methods and adjusts the selected subset throughout the training process. Our proposal has two key components: a) a \textit{coarse-to-fine refinement} schedule for training data, where initial training rounds are performed on larger subsets of data, and the selected subset shrinks throughout the training process, and b) the use of an \textit{unsupervised proxy model} that dynamically selects training instances based on their informativeness for the model’s current state. We also use the proxy model to speed up initial learning by aligning the representations of the primary and proxy models using an additional regularization loss. We validate our method on public benchmarks (CIFAR100, CIFAR10, TinyImagenet, and STL10) and document significant gains in our compute-accuracy tradeoff compared to previous approaches. Notably, we show a 31.6\% reduction in computational load on TinyImagenet while maintaining classification accuracy.
Parameter-Efficient and Memory-Efficient Tuning for Vision Transformer: A Disentangled Approach
PDF ↗Recent works on parameter-efficient transfer learning (PETL) show the potential to adapt a pre-trained Vision Transformer to downstream recognition tasks with only a few learnable parameters. However, since they usually insert new structures into the pre-trained model, entire intermediate features of that model are changed and thus need to be stored to be involved in back-propagation, resulting in memory-heavy training. We solve this problem from a novel disentangled perspective, i.e., dividing PETL into two aspects: task-specific learning and pre-trained knowledge utilization. Specifically, we synthesize the task-specific query with a learnable and lightweight module, which is independent of the pre-trained model. The synthesized query equipped with task-specific knowledge serves to extract the useful features for downstream tasks from the intermediate representations of the pre-trained model in a query-only manner. Built upon these features, a customized classification head is proposed to make the prediction for the input sample. Given that our method employs a extremely lightweight architecture and avoids the use of heavy intermediate features for running gradient descent, it demonstrates limited memory usage in training. Notably, extensive experiments manifest that our method achieves state-of-the-art performance under memory constraints, showcasing its applicability in real-world situations.
Multimodal pre-trained models, such as CLIP, are popular for zero-shot classification due to their open-vocabulary flexibility and high performance. However, vision-language models, which compute similarity scores between images and class labels, are largely black-box, with limited interpretability, risk for bias, and inability to discover new visual concepts not written down. Moreover, in practical settings, the vocabulary for class names and attributes of specialized concepts will not be known, preventing these methods from performing well on images uncommon in large-scale vision-language datasets. To address these limitations, we present a novel method that discovers interpretable yet discriminative sets of attributes for visual recognition. We introduce an evolutionary search algorithm that utilizes a large language model and its in-context learning abilities to iteratively mutate a concept bottleneck of attributes for classification. Our method produces state-of-the-art, interpretable fine-grained classifiers. We outperform the latest baselines by 18.4% on five fine-grained iNaturalist datasets and by 22.2% on two KikiBouba datasets, despite the baselines having access to privileged information about class names.
Addressing fairness in artificial intelligence (AI), particularly in medical AI, is crucial for ensuring equitable healthcare outcomes. Recent efforts to enhance fairness have introduced new methodologies and datasets in medical AI. However, the fairness issue under the setting of domain transfer is almost unexplored, while it is common that clinics rely on different imaging technologies (e.g., different retinal imaging modalities) for patient diagnosis. This paper presents FairDomain, a pioneering systemic study into algorithmic fairness under domain shifts, employing state-of-the-art domain adaptation (DA) and generalization (DG) algorithms for both medical segmentation and classification tasks to understand how biases are transferred between different domains. We also introduce a novel plug-and-play fair identity attention (FIA) module that adapts to various DA and DG algorithms to improve fairness by using self-attention to adjust feature importance based on demographic attributes. Additionally, we curate the first fairness-focused dataset with two paired imaging modalities for the same patient cohort on medical segmentation and classification tasks, to rigorously assess fairness in domain-shift scenarios. Excluding the confounding impact of demographic distribution variation between source and target domains will allow clearer quantification of the performance of domain transfer models. Our extensive evaluations reveal that the proposed FIA significantly enhances both model performance accounted for fairness across all domain shift settings (i.e., DA and DG) with respect to different demographics, which outperforms existing methods on both segmentation and classification. The code and data for this paper can be accessed at https://github.com/anonymous4science/FairDomain.
TTD: Text-Tag Self-Distillation Enhancing Image-Text Alignment in CLIP to Alleviate Single Tag Bias
PDF ↗We identify a critical bias in contemporary CLIP-based models, which we denote as single tag bias. This bias manifests as a disproportionate focus on a singular tag (word) while neglecting other pertinent tags, stemming from CLIP embeddings prioritizing one specific tag in image-text relationships. In this paper, we introduce a novel two-step fine-tuning approach, Text-Tag Self-Distillation (TTD), to address this challenge. We first extract all image-relevant tags from text based on their similarity to the nearest pixels. Then, we distill a combined mask containing the extracted tags' content to a text-derived mask. This approach ensures the unbiased image-text alignment of the CLIP-based models using only image-text pairs without necessitating additional supervision. Our technique demonstrates model-agnostic improvements in multi-tag classification and segmentation tasks, surpassing competing methods that rely on external resources. The code and data are available at https://github.com/shjo-april/TTD.
Understanding human-object interaction (HOI) in videos represents a fundamental yet intricate challenge in computer vision, requiring perception and reasoning across both spatial and temporal domains. Despite previous success of object detection and tracking, multi-person video HOI recognition still faces two major challenges: (1) the three facets of HOI (human, objects, and the interactions that bind them) exhibit interconnectedness and exert mutual influence upon one another. (2) the complexity of multi-person multi-object combinations in spatio-temporal interaction. To address them, we design a spatio-temporal context fuser to better model the interactions among persons and objects in videos. Furthermore, to equip the model with temporal reasoning capacity, we propose an interaction state reasoner module on top of context fuser. Considering the interaction is a key element to bind human and object, we propose an interaction-centric hypersphere in the feature embedding space to model each category of interaction. It helps to learn the distribution of HOI samples belonging to the same interactions on the hypersphere. After training, each interaction prototype sphere will fit the testing HOI sample to determine the HOI classification result. Empirical results on multi-person video HOI dataset MPHOI-72 indicate that our method remarkably surpasses state-of-the-art (SOTA) method by more than 22% F_1 score. At the same time, on single-person datasets Bimanual Actions (single-human two-hand HOI) and CAD-120 (single-human HOI), our method achieves on par or even better results compared with SOTA methods.
The attention module is the key component in Transformers. While the global attention mechanism offers high expressiveness, its excessive computational cost restricts its applicability in various scenarios. In this paper, we propose a novel attention paradigm, Agent Attention, to strike a favorable balance between computational efficiency and representation power. Specifically, the Agent Attention, denoted as a quadruple (Q, A, K, V), introduces an additional set of agent tokens A into the conventional attention module. The agent tokens first act as the agent for the query tokens Q to aggregate information from K and V, and then broadcast the information back to Q. Given the number of agent tokens can be designed to be much smaller than the number of query tokens, agent attention is significantly more efficient than the widely adopted Softmax attention, while preserving global context modelling capability. Interestingly, we show that the proposed agent attention is equivalent to a generalized form of linear attention. Therefore, agent attention seamlessly integrates the powerful Softmax attention and the highly efficient linear attention. Extensive experiments demonstrate the effectiveness of agent attention with various vision Transformers and across diverse vision tasks, including image classification, object detection, semantic segmentation and image generation. Notably, agent attention has shown remarkable performance in high-resolution scenarios, owning to its linear attention nature. For instance, when applied to Stable Diffusion, our agent attention accelerates generation and substantially enhances image generation quality without any additional training. Code will be released.
Remote photoplethysmography (rPPG) is gaining prominence for its non-invasive approach to monitoring physiological signals using only cameras. Despite its promise, the adaptability of rPPG models to new, unseen domains is hindered due to the environmental sensitivity of physiological signals. To address this issue, we pioneer the Test-Time Adaptation (TTA) in rPPG, enabling the adaptation of pre-trained models to the target domain during inference, sidestepping the need for annotations or source data due to privacy considerations. Particularly, utilizing only the user's face video stream as the accessible target domain data, the rPPG model is adjusted by tuning on each single instance it encounters. However, 1) TTA algorithms are designed predominantly for classification tasks, ill-suited in regression tasks such as rPPG due to inadequate supervision. 2) Tuning pre-trained models in a single-instance manner introduces variability and instability, posing challenges to effectively filtering domain-relevant from domain-irrelevant features while simultaneously preserving the learned information. To overcome these challenges, we present \textbf{Bi-TTA}, a novel domain knowledge-based \textbf{Bi}directional \textbf{T}est-\textbf{T}ime \textbf{A}dapter framework. Specifically, leveraging two expert-knowledge priors for providing self-supervision, our Bi-TTA primarily comprises two modules: a prospective adaptation (PA) module using sharpness-aware minimization to eliminate domain-irrelevant noise, enhancing the stability and efficacy during the adaptation process, and a retrospective stabilization (RS) module to dynamically reinforce crucial learned model parameters, averting performance degradation caused by overfitting or catastrophic forgetting. To this end, we established a large-scale benchmark for rPPG tasks under TTA protocol, promoting advancements in both the rPPG and TTA fields. The experimental results demonstrate the significant superiority of our approach over the state-of-the-art (SoTA).
Modern pre-trained architectures struggle to retain previous information while undergoing continuous fine-tuning on new tasks. Despite notable progress in continual classification, systems designed for complex vision tasks such as detection or segmentation still struggle to attain satisfactory performance. In this work, we introduce a memory-based detection transformer architecture to adapt a pre-trained DETR-style detector to new tasks while preserving knowledge from previous tasks. We propose a novel localized query function for efficient information retrieval from memory units, aiming to minimize forgetting. Furthermore, we identify a fundamental challenge in continual detection referred to as {\em background relegation}. This arises when object categories from earlier tasks reappear in future tasks, potentially without labels, leading them to be implicitly treated as background. This is an inevitable issue in continual detection or segmentation. The introduced continual optimization technique effectively tackles this challenge. Finally, we assess the performance of our proposed system on continual detection benchmarks and demonstrate that our approach surpasses the performance of existing state-of-the-art resulting in 5-7\% improvements on MSCOCO and PASCAL-VOC on the task of continual detection.
Localization and Expansion: A Decoupled Framework for Point Cloud Few-shot Semantic Segmentation
PDF ↗Point cloud few-shot semantic segmentation (PC-FSS) aims to segment targets of novel categories in a given query point cloud with only a few annotated support samples. The current top-performing prototypical learning methods employ prototypes originating from support samples to direct the classification of query points. However, the inherent fragility of point-level matching and the prevalent intra-class diversity pose great challenges to this cross-instance matching paradigm, leading to erroneous background activations or incomplete target excavation. In this work, we propose a simple yet effective framework in the spirit of Decoupled Localization and Expansion (DLE). The proposed DLE, including a structural localization module (SLM) and a self-expansion module (SEM), enjoys several merits. First, structural information is injected into the matching process through the agent-level correlation in SLM, and the confident target region can thus be precisely located. Second, more reliable intra-object similarity is harnessed in SEM to derive the complete target, and the conservative expansion strategy is introduced to reasonably constrain the expansion. Extensive experiments on two challenging benchmarks under different settings demonstrate that DLE outperforms previous state-of-the-art approaches by large margins.
Monocular depth estimation is a challenging task that predicts the pixel-wise depth from a single 2D image. Current methods typically model this problem as a regression or classification task. We propose DiffusionDepth, a new approach that reformulates monocular depth estimation as a denoising diffusion process. It learns an iterative denoising process to `denoise' random depth distribution into a depth map with the guidance of monocular visual conditions. The process is performed in the latent space encoded by a dedicated depth encoder and decoder. Instead of diffusing ground truth (GT) depth, the model learns to reverse the process of diffusing the refined depth of itself into random depth distribution. This self-diffusion formulation overcomes the difficulty of applying generative models to sparse GT depth scenarios. The proposed approach benefits this task by refining depth estimation step by step, which is superior for generating accurate and highly detailed depth maps. Experimental results on KITTI and NYU-Depth-V2 datasets suggest that a simple yet efficient diffusion approach could reach state-of-the-art performance in both indoor and outdoor scenarios with acceptable inference time.
Discriminative models for object classification typically learn image-based representations that do not capture the compositional and 3D nature of objects. In this work, we show that explicitly integrating 3D compositional object representations into deep networks for image classification leads to a largely enhanced generalization in out-of-distribution scenarios. In particular, we introduce a novel architecture, referred to as NOVUM, that consists of a feature extractor and a neural object volume for every target object class. Each neural object volume is a composition of 3D Gaussians that emit feature vectors. This compositional object representation allows for a highly robust and fast estimation of the object class by independently matching the features of the 3D Gaussians of each category to features extracted from an input image. Additionally, the object pose can be estimated via inverse rendering of the corresponding neural object volume. To enable the classification of objects, the neural features at each 3D Gaussian are trained discriminatively to be distinct from (i) the features of 3D Gaussians in other categories, (ii) features of other 3D Gaussians of the same object, and (iii) the background features. Our experiments show that NOVUM offers intriguing advantages over standard architectures due to the 3D compositional structure of the object representation, namely: (1) An exceptional robustness across a spectrum of real-world and synthetic out-of-distribution shifts and (2) an enhanced human interpretability compared to standard models, all while maintaining real-time inference and a competitive accuracy on in-distribution data.
In the realm of Adversarial Distillation (AD), strategic and precise knowledge transfer from an adversarially robust teacher model to a less robust student model is paramount. Our Dynamic Guidance Adversarial Distillation (DGAD) framework directly tackles the challenge of differential sample importance, with a keen focus on rectifying the teacher model's misclassifications. DGAD employs Misclassification-Aware Partitioning (MAP) to dynamically tailor the distillation focus, optimizing the learning process by steering towards the most reliable teacher predictions. Additionally, our Error-corrective Label Swapping (ELS) corrects teacher's misclassifications on both clean and adversarially perturbed inputs, refining the quality of knowledge transfer. Further, Predictive Consistency Regularization (PCR) guarantees consistent performance of the student model across both clean and adversarial inputs, significantly enhancing its overall robustness. By integrating these methodologies, DGAD significantly improves upon the accuracy of clean data and fortifies the model's defenses against sophisticated adversarial threats. Our experimental validation on CIFAR10, CIFAR100, and Tiny ImageNet datasets, employing various model architectures, demonstrates the efficacy of DGAD, establishing it as a promising approach for enhancing both the robustness and accuracy of student models in adversarial settings.
We present a new open-vocabulary detection approach based on region-centric image-language pretraining to bridge the gap between image-level pretraining and open-vocabulary object detection. At the pretraining phase, we incorporate the detector architecture on top of the classification backbone, which better serves the region-level recognition needs of detection by enabling the detector heads to learn from large-scale image-text pairs. Using only standard contrastive loss and no pseudo-labeling, our approach is a simple yet effective extension of the contrastive learning method to learn emergent object-semantic cues. In addition, we propose a shifted-window learning approach upon window attention to make the backbone representation more robust, translation-invariant, and less biased by the window pattern. On the popular LVIS open-vocabulary detection benchmark, our approach sets a new state of the art of 37.6 mask APr using the common ViT-L backbone and public LAION dataset, significantly outperforming the best existing approach by +3.7 mask APr at system level. On the COCO benchmark, we achieve very competitive 39.6 novel AP without pseudo labeling or weak supervision. In addition, we evaluate our approach on the transfer detection setup, where it demonstrates notable improvement over the baseline. Visualization reveals emerging object locality from the pretraining recipes compared to the baseline. Code and models will be publicly released.