论文检索

Chenxu Wang, Chunyan Xu, Xiang Li, YuXuan Li, Xu Guo, Ziqi Gu, Zhen Cui

While existing semi-supervised object detection (SSOD) methods perform well in general scenes, they encounter challenges in handling oriented objects in aerial images. We experimentally find three gaps between general and oriented object detection in semi-supervised learning: 1) Sampling inconsistency: the common center sampling is not suitable for oriented objects with larger aspect ratios when selecting positive labels from labeled data. 2) Assignment inconsistency: balancing the precision and localization quality of oriented pseudo-boxes poses greater challenges which introduces more noise when selecting positive labels from unlabeled data. 3) Confidence inconsistency: there exists more mismatch between the predicted classification and localization qualities when considering oriented objects, affecting the selection of pseudo-labels. Therefore, we propose a Multi-clue Consistency Learning (MCL) framework to bridge gaps between general and oriented objects in semi-supervised detection. Specifically, considering various shapes of rotated objects, the Gaussian Center Assignment is specially designed to select the pixel-level positive labels from labeled data. We then introduce the Scale-aware Label Assignment to select pixel-level pseudo-labels instead of unreliable pseudo-boxes, which is a divide-and-rule strategy suited for objects with various scales. The Consistent Confidence Soft Label is adopted to further boost the detector by maintaining the alignment of the predicted results. Comprehensive experiments on DOTA-v1.5 and DOTA-v1.0 benchmarks demonstrate that our proposed MCL can achieve state-of-the-art performance in the semi-supervised oriented object detection task.

Bo Wang, Zhao Zhang, Suiyi Zhao, Xianming Ye, Haijun Zhang, Meng Wang

Existing face forgery detection methods achieve promising performance when training and testing forgery data are from identical manipulation types, while they fail to generalize well to unseen samples. In this paper, we experimentally investigate and find that the poor generalization of the methods mainly arises from their overfitting on the known fake patterns. Excessively focused on seen fakes, those detectors fail to effectively learn image-intrinsic information and the distributional disparity between real and fake images. Then, to address this issue, we redefine fake learning as real-fake distributional disparity learning. We propose a novel deepfake detection framework learning distributional disparity based on the differentiated reconstruction on real and fake images for improved generalization. Specifically, distributional disparity learning on differentiated reconstruction of the real and fake images, enforces the model to learn image-invariant intrinsic representations. The reconstruction on real and fake images forces the decoders to learn the distribution of real and fake images, respectively. Moreover, to avoid the influence from the specificalization of the known fake patterns, we further propose the information interaction learning on the encoded intrinsic information and the pixel disparity between the input image and its reconstruction to distinguish face forgeries that are even unknown. Extensive experiments on large-scale benchmark datasets demonstrated the effectiveness of addressing the overfitting issue of the classification network, and verified the superior performance of our method.

Aowen Wang, Zhiwang Zhang, Dongang Wang, Fanyi Wang, Haotian Hu, Jinyang Guo, Yipeng Zhou, Chaoyi Pang, Shiting Wen

The scarcity data of medical field brings the collaborative training in medical vision-language pre-training (VLP) cross different clients. Therefore, the collaborative training in medical VLP faces two challenges: First, the medical data requires privacy, thus can not directly shared across different clients. Second, medical data distribution across institutes is typically heterogeneous, hindering local model alignment and representation capabilities. To simultaneously overcome these two challenges, we propose the framework called personalized model selector with fused multimodal information (PMS-FM). The contribution of PMS-FM is two-fold: 1) PMS-FM uses embeddings to represent information in different formats, allowing for the fusion of multimodal data. 2) PMS-FM adapts to personalized data distributions by training multiple models. A model selector then identifies and selects the best-performing model for each individual client. Extensive experiments with multiple real-world medical datasets demonstrate the superb performance of PMS-FM over existing federated learning methods on different zero-shot classification tasks.

Haitao Tian, Pierre Payeur

Existing skeleton-based human action classification models rely on well-trimmed action-specific skeleton videos for both training and testing, precluding their scalability to real-world applications where untrimmed videos exhibiting concatenated actions are predominant. To overcome this limitation, recently introduced skeleton action segmentation models involve un-trimmed skeleton videos into end-to-end training. The model is optimized to provide frame-wise predictions for any length of testing videos, simultaneously realizing action localization and classification. Yet, achieving such an improvement im-poses frame-wise annotated skeleton videos, which remains time-consuming in practice. This paper features a novel framework for skeleton-based action segmentation trained on short trimmed skeleton videos, but that can run on longer un-trimmed videos. The approach is implemented in three steps: Stitch, Contrast, and Segment. First, Stitch proposes a tem-poral skeleton stitching scheme that treats trimmed skeleton videos as elementary human motions that compose a semantic space and can be sampled to generate multi-action stitched se-quences. Contrast learns contrastive representations from stitched sequences with a novel discrimination pretext task that enables a skeleton encoder to learn meaningful action-temporal contexts to improve action segmentation. Finally, Segment relates the proposed method to action segmentation by learning a segmentation layer while handling particular da-ta availability. Experiments involve a trimmed source dataset and an untrimmed target dataset in an adaptation formulation for real-world skeleton-based human action segmentation to evaluate the effectiveness of the proposed method.

Yuchen Su, Zhineng Chen, Yongkun Du, Zhilong Ji, Kai Hu, Jinfeng Bai, Xieping Gao

Connected component (CC) is a proper text shape representation that aligns with human reading intuition. However, CC-based text detection methods have recently faced a developmental bottleneck that their time-consuming post-processing is difficult to eliminate. To address this issue, we introduce an explicit relational reasoning network (ERRNet) to elegantly model the component relationships without post-processing. Concretely, we first represent each text instance as multiple ordered text components, and then treat these components as objects in sequential movement. In this way, scene text detection can be innovatively viewed as a tracking problem. From this perspective, we design an end-to-end tracking decoder to achieve a CC-based method dispensing with post-processing entirely. Additionally, we observe that there is an inconsistency between classification confidence and localization quality, so we propose a Polygon Monte-Carlo method to quickly and accurately evaluate the localization quality. Based on this, we introduce a position-supervised classification loss to guide the task-aligned learning of ERRNet. Experiments on challenging benchmarks demonstrate the effectiveness of our ERRNet. It consistently achieves state-of-the-art accuracy while holding highly competitive inference speed.

Yaling Shen, Zhixiong Zhuang, Kun Yuan, Maria-Irina Nicolae, Nassir Navab, Nicolas Padoy, Mario Fritz

Medical multimodal large language models (MLLMs) are becoming an instrumental part of healthcare systems, assisting medical personnel with decision making and results analysis. Models for radiology report generation are able to interpret medical imagery, thus reducing the workload of radiologists. As medical data is scarce and protected by privacy regulations, medical MLLMs represent valuable intellectual property. However, these assets are potentially vulnerable to model stealing, where attackers aim to replicate their functionality via black-box access. So far, model stealing for the medical domain has focused on image classification; however, existing attacks are not effective against MLLMs. In this paper, we introduce Adversarial Domain Alignment (ADA-Steal), the first stealing attack against medical MLLMs. ADA-Steal relies on natural images, which are public and widely available, as opposed to their medical counterparts. We show that data augmentation with adversarial noise is sufficient to overcome the data distribution gap between natural images and the domain-specific distribution of the victim MLLM. Experiments on the IU X-RAY and MIMIC-CXR radiology datasets demonstrate that Adversarial Domain Alignment enables attackers to steal the medical MLLM without any access to medical data.

Joon Hyun Park, Kumju Jo, Sungyong Baik

Entrusted with the goal of pixel-level object classification, the semantic segmentation networks entails the laborious preparation of pixel-level annotation masks. To obtain pixel-level annotation masks for a given class without human efforts, recent few works have proposed to generate pairs of images and annotation masks by employing image and text relationships modeled by text-to-image generative models, especially Stable Diffusion. However, these works do not fully exploit the capability of text-guided Diffusion models and thus require a pre-trained segmentation network, careful text prompt tuning, or the training of a segmentation network to generate final annotation masks. In this work, we take a closer look at attention mechanisms of Stable Diffusion, from which we draw connections with classical seeded segmentation approaches. In particular, we show that cross-attention alone provides very coarse object localization, which however can provide initial seeds. Then, akin to region expansion in seeded segmentation, we utilize the semantic-correspondence-modeling capability of self-attention to iteratively spread the attention to the whole class from the seeds using multi-scale self-attention maps. We also observe that a simple-text-guided synthetic image often has a uniform background, which is easier to find correspondences, compared to complex-structured objects. Thus, we further refine a mask using a more accurate background mask. Our proposed method, dubbed SeeDiff, generates high-quality masks off-the-shelf from Stable Diffusion, without additional training procedure, prompt tuning, or a pre-trained segmentation network.

Chao Pang, Xingxing Weng, Jiang Wu, Jiayu Li, Yi Liu, Jiaxing Sun, Weijia Li, Shuai Wang, Litong Feng, Gui-Song Xia 等

This paper develops a Versatile and Honest vision language Model (VHM) for remote sensing image analysis. VHM is built on a large-scale remote sensing image-text dataset with rich-content captions (VersaD), and an honest instruction dataset comprising both factual and deceptive questions (HnstD). Unlike prevailing remote sensing image-text datasets, in which image captions focus on a few prominent objects and their relationships, VersaD captions provide detailed information about image properties, object attributes, and the overall scene. This comprehensive captioning enables VHM to thoroughly understand remote sensing images and perform diverse remote sensing tasks. Moreover, different from existing remote sensing instruction datasets that only include factual questions, HnstD contains additional deceptive questions stemming from the non-existence of objects. This feature prevents VHM from producing affirmative answers to nonsense queries, thereby ensuring its honesty. In our experiments, VHM significantly outperforms various vision language models on common tasks of scene classification, visual question answering, and visual grounding. Additionally, VHM achieves competent performance on several unexplored tasks, such as building vectorizing, multi-label classification and honest question answering.

Subhodip Panda, Shashwat Sourav, Prathosh A.P.

To follow regulations on individual data privacy and safety, machine learning models must systematically remove information learned from specific subsets of a user's training data that can no longer be utilized. To address this problem, machine unlearning has emerged as an important area of research, that helps remove information learned from specific subsets of training data from a pre-trained model without needing to retrain the whole model from scratch. The principal aim of this study is to formulate a methodology aimed for the purposeful elimination of information linked to a specific class of data from a pre-trained classification network. This intentional removal decreases the model's performance specifically concerning the unlearned data class while simultaneously minimizing any detrimental impacts on the model's performance in other classes. To achieve this goal, we frame the class unlearning problem from a Bayesian perspective, which yields a loss function that minimizes the log-likelihood associated with the unlearned data with a stability regularization in parameter space. This stability regularization incorporates Mohalanobis distance with respect to the Fisher Information matrix and L2 distance from the pre-trained model parameters. Our novel approach, termed Partially-Blinded Unlearning (PBU), surpasses existing state-of-the-art class unlearning methods, demonstrating superior effectiveness. Notably, PBU achieves this efficacy without requiring information about the entire training dataset but only of the unlearned data points, marking a distinctive feature of its performance.

Xiaotian Pan, Zhaobo Qi, Xin Sun, Yuanrong Xu, Weigang Zhang

Procedure planning in instructional videos, producing a structured and plannable action sequence facilitating the transition from the start to the goal states, has achieved significant progress. The dominant single-branch non-autoregressive planning paradigm guides action sequence generation through action labels, overlooking the limitation of the absence of intermediate visual information. Hence, we introduce the procedure knowledge decoupled distillation strategy to address the above issue. This innovative strategy deliberately lets the teacher model see the real visual information among the start and goal states to enhance its action semantic understanding and relationship modeling ability, producing the potential probability distribution containing the real action class and other action classes that may occur. Accordingly, we introduce a decoupled intermediate information knowledge distillation loss, which comprises single action knowledge distillation and sequence distribution knowledge distillation for the student model. The former improves the student model's precise inference ability for individual actions by transferring knowledge of a single action target category using binary classification loss. Conversely, the latter uses MSE loss to constrain the student model to learn the action sequence probability distribution from the teacher model, thereby enhancing the student model's global planning capability. Extensive experiments on three datasets demonstrate that our strategy can improve the performance of multiple weakly supervised models, achieving promising procedure knowledge modeling ability and plug-and-play flexibility.

Wenjie Pan, Jianqing Zhu, Huanqiang Zeng

There are two manifestations of classification fairness. One is the preference for head classes with more instances due to the long-tail (LT) distribution of training data. The other is the clever Hans (CH) effect, where non-discriminative features are mistakenly used for classification. In this paper, we find that using category-agnostic zero-valued data can simultaneously reveal both types of unfairness. Based on this, we propose a zero uniformity training (ZUT) framework to optimize classification fairness. The ZUT framework inputs category-agnostic zero-valued data into the model in parallel and uses zero uniformity loss (ZUL) to optimize classification fairness. The ZUL loss mitigates bias towards specific classes by unifying the classification features corresponding to zero-valued data. The ZUT framework is compatible with various classification-based tasks. Experiments show that the ZUT framework can improve the performance of multiple state-of-the-art methods in image classification, person re-identification, and semantic segmentation.

Hongwei Niu, Linhuang Xie, Jianghang Lin, Shengchuan Zhang

Domain Generalized Semantic Segmentation (DGSS) seeks to utilize source domain data exclusively to enhance the generalization of semantic segmentation across unknown target domains. Prevailing studies predominantly concentrate on feature normalization and domain randomization, these approaches exhibit significant limitations. Feature normalization-based methods tend to confuse semantic features in the process of constraining the feature space distribution, resulting in classification misjudgment. Domain randomization-based methods frequently incorporate domain-irrelevant noise due to the uncontrollability of style transformations, resulting in segmentation ambiguity. To address these challenges, we introduce a novel framework, named SCSD for Semantic Consistency prediction and Style Diversity generalization. It comprises three pivotal components: Firstly, a Semantic Query Booster is designed to enhance the semantic awareness and discrimination capabilities of object queries in the mask decoder, enabling cross-domain semantic consistency prediction. Secondly, we develop a Text-Driven Style Transform module that utilizes domain difference text embeddings to controllably guide the style transformation of image features, thereby increasing inter-domain style diversity. Lastly, to prevent the collapse of similar domain feature spaces, we introduce a Style Synergy Optimization mechanism that fortifies the separation of inter-domain features and the aggregation of intra-domain features by synergistically weighting style contrastive loss and style aggregation loss. Extensive experiments demonstrate that the proposed SCSD significantly outperforms existing state-of-theart methods. Notably, SCSD trained on GTAV achieved an average of 49.11 mIoU on the four unseen domain datasets, surpassing the state-of-the-art method by +4.08 mIoU.

Ba Hung Ngo, Doanh C. Bui, Nhat-Tuong Do-Tran, Tae Jong Choi

The enhanced representational power and broad applicability of deep learning models have attracted significant interest from the research community in recent years. However, these models often struggle to perform effectively under domain shift conditions, where the training data (the source domain) is related to but exhibits different distributions from the testing data (the target domain). To address this challenge, previous studies have attempted to reduce the domain gap between source and target data by incorporating a few labeled target samples during training—a technique known as semi-supervised domain adaptation (SSDA). While this strategy has demonstrated notable improvements in classification performance, the network architectures used in these approaches primarily focus on exploiting the features of individual images, leaving room for improvement in capturing rich representations. In this study, we introduce a Hierarchical Graph of Nodes designed to simultaneously present representations at both feature and category levels. At the feature level, we introduce a local graph to identify the most relevant patches within an image, facilitating adaptability to defined main object representations. At the category level, we employ a global graph to aggregate the features from samples within the same category, thereby enriching overall representations. Extensive experiments on widely used SSDA benchmark datasets, including Office-Home, DomainNet, and VisDA2017, demonstrate that both quantitative and qualitative results substantiate the effectiveness of HiGDA, establishing it as a new state-of-the-art method.

Hefei Mei, Minjing Dong, Chang Xu

Diffusion models (DMs) have demonstrated great potential in the field of adversarial robustness, where DM-based defense methods can achieve superior defense capability without adversarial training. However, they all require huge computational costs due to the usage of large-scale pre-trained DMs, making it difficult to conduct full evaluation under strong attacks and compare with traditional CNN-based methods. Simply reducing the network size and timesteps in DMs could significantly harm the image generation quality, which invalidates previous frameworks. To alleviate this issue, we redesign the diffusion framework from generating high-quality images to predicting distinguishable image labels. Specifically, we employ an image translation framework to learn many-to-one mapping from input samples to designed orthogonal image labels. Based on this framework, we introduce an efficient Image-to-Image diffusion classifier with a pruned U-Net structure and reduced diffusion timesteps. Besides the framework, we redesign the optimization objective of DMs to fit the target of image classification, where a new classification loss is incorporated in the DM-based image translation framework to distinguish the generated label from those of other classes. We conduct sufficient evaluations of the proposed classifier under various attacks on popular benchmarks. Extensive experiments show that our method achieves better adversarial robustness with fewer computational costs than DM-based and CNN-based methods.

Zhen-Xiang Ma, Zhen-Duo Chen, Tai Zheng, Xin Luo, Zixia Jia, Xin-Shun Xu

Recently, a number of effective methods have been proposed to tackle the challenging task of Few-Shot Fine-Grained Image Classification (FS-FGIC). However, how to fully leverage the backbone network to discover and extract detailed features to generate more discriminative class prototypes, as well as how to accurately model the similarity relationship between query samples and the class prototypes, are still issues to be further considered. Therefore, we propose a novel progreSsively featUre refInement and conTinuous rElationship moDeling method, SUITED for short, to address these two issues existing in the State-of-the-Art FS-FGIC methods. Specifically, we design the Progressive Feature Refinement Module (PFRM) to fully exploit the backbone network's progressive feature extraction capabilities, forming multi-scale feature representations to further enhance discriminative features. Then, the Continuous Relationship Modeling Module (CRMM) is proposed to capture the dependencies between query samples and the corresponding class prototypes, achieving precise optimization of the distances among corresponding sample points in the feature space. We conducted extensive experiments on five fine-grained benchmark datasets, and the experimental results demonstrate that the proposed method is comprehensively ahead of the existing State-of-the-Art methods.

Weijian Ma, Ruoxin Chen, Keyue Zhang, Shuang Wu, Shouhong Ding

Data augmentation is expected to bring about unseen features of training set, enhancing the model’s ability to generalize in situations where data is limited. Generative image models trained on large web-crawled datasets such as LAION are known to produce images with stereotypes and imperceptible bias when used to augment training data, owing to dataset misalignment and the generator’s ignorance of the downstream model. We improve downstream task awareness in generated images by proposing a task-aware fine-tuning strategy that actively detects failures of downstream task in the target model to fine-tune the generation process between epochs. The dynamic fine-tuning strategy is achieved by (1) inspecting misalignment between generated data and original data via VLM captioners and (2) adjusts both prompts and diffusion model so that the strategy dynamically guides the generator by focusing on the detected bias of VLM. This is done via re-captioning the overfitted data as well as finetuning the diffusion trajectory in a contrastive manner. To co-operate with the VLM captioner, the contrastive fine-tuning process dynamically adjusts different parts of the diffusion trajectory based on detected misalignment, thus shifting the the generated distribution away from making the downstream model overfit. Our experiments on few-shot class incremental learning show that our instruction-guided finetuning strategy consistently assists the downstream model with higher classification accuracy compared to generative data augmentation baselines such as Stable Diffusion and GPT-4o, and state-of-the-art non-generative strategies.

Pingchuan Ma, Lennart Rietdorf, Dmytro Kotovenko, Vincent Tao Hu, Björn Ommer

Accurately describing images with text is a foundation of explainable AI. Vision-Language Models (VLMs) like CLIP have recently addressed this by aligning images and texts in a shared embedding space, expressing semantic similarities between vision and language embeddings. VLM classification can be improved with descriptions generated by Large Language Models (LLMs). However, it is difficult to determine the contribution of actual description semantics, as the performance gain may also stem from a semantic-agnostic ensembling effect, where multiple modified text prompts act as a noisy test-time augmentation for the original one. We propose an alternative evaluation scenario to decide if a performance boost of LLM-generated descriptions is caused by such a noise augmentation effect or rather by genuine description semantics. The proposed scenario avoids noisy test-time augmentation and ensures that genuine, distinctive descriptions cause the performance boost. Furthermore, we propose a training-free method for selecting discriminative descriptions that work independently of classname-ensembling effects. Our approach identifies descriptions that effectively differentiate classes within a local CLIP label neighborhood, improving classification accuracy across seven datasets. Additionally, we provide insights into the explainability of description-based image classification using VLMs.

Jitao Ma, Weiying Xie, Hangyu Ye, Daixun Li, Leyuan Fang

Zero-shot generalized anomaly detection (ZGAD) plays a critical role in industrial automation and health screening. Recent studies have shown that ZGAD methods built on visual-language models (VLMs) like CLIP have excellent cross-domain detection performance. Different from other computer vision tasks, ZGAD needs to jointly optimize both image-level anomaly classification and pixel-level anomaly segmentation tasks for determining whether an image contains anomalies and detecting anomalous parts of an image, respectively, this leads to different granularity of the tasks. However, existing methods ignore this problem, processing these two tasks with one set of broad text prompts used to describe the whole image. This limits CLIP to align textual features with pixel-level visual features and impairs anomaly segmentation performance. Therefore, for precise visual-text alignment, in this paper we propose a novel fine-grained text prompts generation strategy. We then apply the broad text prompts and the generated fine-grained text prompts for visual-textual alignment in classification and segmentation tasks, respectively, accurately capturing normal and anomalous instances in images. We also introduce the Text Prompt Shunt (TPS) model, which performs joint learning by reconstruction the complementary and dependency relationships between the two tasks to enhance anomaly detection performance. This enables our method to focus on fine-grained segmentation of anomalous targets while ensuring accurate anomaly classification, and achieve pixel-level comprehensible CLIP for the first time in the ZGAD task. Extensive experiments on 13 real-world anomaly detection datasets demonstrate that TPS achieves superior ZGAD performance across highly diverse datasets from industrial and medical domains.

Shuai Lyu, Rongchen Zhang, Zeqi Ma, Fangjian Liao, Dongmei Mo, Waikeung Wong

Few-shot defect multi-classification (FSDMC) is an emerging trend in quality control within industrial manufacturing. However, current FSDMC research often lacks generalizability due to its focus on specific datasets. Additionally, defect classification heavily relies on contextual information within images, and existing methods fall short of effectively extracting this information. To address these challenges, we propose a general FSDMC framework called MVREC, which offers two primary advantages: (1) MVREC extracts general features for defect instances by incorporating the pre-trained AlphaCLIP model. (2) It utilizes a region-context framework to enhance defect features by leveraging mask region input and multi-view context augmentation. Furthermore, Few-shot Zip-Adapter(-F) classifiers within the model are introduced to cache the visual features of the support set and perform few-shot classification. We also introduce MVTec-FS, a new FSDMC benchmark based on MVTec AD, which includes 1228 defect images with instance-level mask annotations and 46 defect types. Extensive experiments conducted on MVTec-FS and four additional datasets demonstrate its effectiveness in general defect classification and its ability to incorporate contextual information to improve classification performance.

HaoMiao Liu, Hao Xu, Chuhuai Yue, Bo Ma

Unknown Object Detection (UOD) aims to identify objects of unseen categories, differing from the traditional detection paradigm limited by the closed-world assumption. A key component of UOD is learning a generalized representation, i.e. objectness for both known and unknown categories to distinguish and localize objects from the background in a class-agnostic manner. However, previous methods obtain supervision signals for learning objectness in isolation from either localization or classification information, leading to poor performance for UOD. To address this issue, we propose a transformer-based UOD framework, UN-DETR. Based on this, we craft Instance Presence Score (IPS) to represent the probability of an object's presence. For the purpose of information complementarity, IPS employs a strategy of joint supervised learning, integrating attributes representing general objectness from the positional and the categorical latent space as supervision signals. To enhance IPS learning, we introduce a one-to-many assignment strategy to incorporate more supervision. Then, we propose Unbiased Query Selection to provide premium initial query vectors for the decoder. Additionally, we propose an IPS-guided post process strategy to filter redundant boxes and correct classification predictions for known and unknown objects. Finally, we pretrain the entire UN-DETR in an unsupervised manner, in order to obtain objectness prior. Our UN-DETR is comprehensively evaluated on multiple UOD and known detection benchmarks, demonstrating its effectiveness and achieving state-of-the-art performance.