PyTorchKR Papers

최신 논문

총 1,759편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2410.13010

Hiding-in-Plain-Sight (HiPS) Attack on CLIP for Targetted Object Removal from Images

Machine learning models are known to be vulnerable to adversarial attacks, but traditional attacks have mostly focused on single-modalities. With the rise of large multi-modal models (LMMs) like CLIP, which combine vision and language capabilities, new vulnerabilities have emerged. However, prior work in multimodal targeted attacks aim to completely change the model's output to what the adversary wants. In many realistic scenarios, an adversary might seek to make only subtle modifications to the output, so that the changes go unnoticed by downstream models or even by humans. We introduce Hiding-in-Plain-Sight (HiPS) attacks, a novel class of adversarial attacks that subtly modifies model predictions by selectively concealing target object(s), as if the target object was absent from the scene. We propose two HiPS attack variants, HiPS-cls and HiPS-cap, and demonstrate their effectiveness in transferring to downstream image captioning models, such as CLIP-Cap, for targeted object removal from image captions.

arXiv:2410.13360공식 코드 · Papers with Code

RAP: Retrieval-Augmented Personalization for Multimodal Large Language Models

The development of large language models (LLMs) has significantly enhanced the capabilities of multimodal LLMs (MLLMs) as general assistants. However, lack of user-specific knowledge still restricts their application in human's daily life. In this paper, we introduce the Retrieval Augmented Personalization (RAP) framework for MLLMs' personalization. Starting from a general MLLM, we turn it into a personalized assistant in three steps. (a) Remember: We design a key-value database to store user-related information, e.g., user's name, avatar and other attributes. (b) Retrieve: When the user initiates a conversation, RAP will retrieve relevant information from the database using a multimodal retriever. (c) Generate: The input query and retrieved concepts' information are fed into MLLMs to generate personalized, knowledge-augmented responses. Unlike previous methods, RAP allows real-time concept editing via updating the external database. To further improve generation quality and alignment with user-specific information, we design a pipeline for data collection and create a specialized dataset for personalized training of MLLMs. Based on the dataset, we train a series of MLLMs as personalized multimodal assistants. By pretraining on large-scale dataset, RAP-MLLMs can generalize to infinite visual concepts without additional finetuning. Our models demonstrate outstanding flexibility and generation quality across a variety of tasks, such as personalized image captioning, question answering and visual recognition. The code, data and models are available at https://hoar012.github.io/RAP-Project/.

arXiv:2410.11321공식 코드 · Papers with Code

Self-adaptive Multimodal Retrieval-Augmented Generation

Traditional Retrieval-Augmented Generation (RAG) methods are limited by their reliance on a fixed number of retrieved documents, often resulting in incomplete or noisy information that undermines task performance. Although recent adaptive approaches alleviated these problems, their application in intricate and real-world multimodal tasks remains limited. To address these, we propose a new approach called Self-adaptive Multimodal Retrieval-Augmented Generation (SAM-RAG), tailored specifically for multimodal contexts. SAM-RAG not only dynamically filters relevant documents based on the input query, including image captions when needed, but also verifies the quality of both the retrieved documents and the output. Extensive experimental results show that SAM-RAG surpasses existing state-of-the-art methods in both retrieval accuracy and response generation. By further ablation experiments and effectiveness analysis, SAM-RAG maintains high recall quality while improving overall task performance in multimodal RAG task. Our codes are available at https://github.com/SAM-RAG/SAM_RAG.

arXiv:2410.10407

MMCFND: Multimodal Multilingual Caption-aware Fake News Detection for Low-resource Indic Languages

The widespread dissemination of false information through manipulative tactics that combine deceptive text and images threatens the integrity of reliable sources of information. While there has been research on detecting fake news in high resource languages using multimodal approaches, methods for low resource Indic languages primarily rely on textual analysis. This difference highlights the need for robust methods that specifically address multimodal fake news in Indic languages, where the lack of extensive datasets and tools presents a significant obstacle to progress. To this end, we introduce the Multimodal Multilingual dataset for Indic Fake News Detection (MMIFND). This meticulously curated dataset consists of 28,085 instances distributed across Hindi, Bengali, Marathi, Malayalam, Tamil, Gujarati and Punjabi. We further propose the Multimodal Multilingual Caption-aware framework for Fake News Detection (MMCFND). MMCFND utilizes pre-trained unimodal encoders and pairwise encoders from a foundational model that aligns vision and language, allowing for extracting deep representations from visual and textual components of news articles. The multimodal fusion encoder in the foundational model integrates text and image representations derived from its pairwise encoders to generate a comprehensive cross modal representation. Furthermore, we generate descriptive image captions that provide additional context to detect inconsistencies and manipulations. The retrieved features are then fused and fed into a classifier to determine the authenticity of news articles. The curated dataset can potentially accelerate research and development in low resource environments significantly. Thorough experimentation on MMIFND demonstrates that our proposed framework outperforms established methods for extracting relevant fake news detection features.

arXiv:2410.09382

CLIP-SCGI: Synthesized Caption-Guided Inversion for Person Re-Identification

Person re-identification (ReID) has recently benefited from large pretrained vision-language models such as Contrastive Language-Image Pre-Training (CLIP). However, the absence of concrete descriptions necessitates the use of implicit text embeddings, which demand complicated and inefficient training strategies. To address this issue, we first propose one straightforward solution by leveraging existing image captioning models to generate pseudo captions for person images, and thereby boost person re-identification with large vision language models. Using models like the Large Language and Vision Assistant (LLAVA), we generate high-quality captions based on fixed templates that capture key semantic attributes such as gender, clothing, and age. By augmenting ReID training sets from uni-modality (image) to bi-modality (image and text), we introduce CLIP-SCGI, a simple yet effective framework that leverages synthesized captions to guide the learning of discriminative and robust representations. Built on CLIP, CLIP-SCGI fuses image and text embeddings through two modules to enhance the training process. To address quality issues in generated captions, we introduce a caption-guided inversion module that captures semantic attributes from images by converting relevant visual information into pseudo-word tokens based on the descriptions. This approach helps the model better capture key information and focus on relevant regions. The extracted features are then utilized in a cross-modal fusion module, guiding the model to focus on regions semantically consistent with the caption, thereby facilitating the optimization of the visual encoder to extract discriminative and robust representations. Extensive experiments on four popular ReID benchmarks demonstrate that CLIP-SCGI outperforms the state-of-the-art by a significant margin.

arXiv:2410.07593공식 코드 · Papers with Code

A Unified Debiasing Approach for Vision-Language Models across Modalities and Tasks

Recent advancements in Vision-Language Models (VLMs) have enabled complex multimodal tasks by processing text and image data simultaneously, significantly enhancing the field of artificial intelligence. However, these models often exhibit biases that can skew outputs towards societal stereotypes, thus necessitating debiasing strategies. Existing debiasing methods focus narrowly on specific modalities or tasks, and require extensive retraining. To address these limitations, this paper introduces Selective Feature Imputation for Debiasing (SFID), a novel methodology that integrates feature pruning and low confidence imputation (LCI) to effectively reduce biases in VLMs. SFID is versatile, maintaining the semantic integrity of outputs and costly effective by eliminating the need for retraining. Our experimental results demonstrate SFID's effectiveness across various VLMs tasks including zero-shot classification, text-to-image retrieval, image captioning, and text-to-image generation, by significantly reducing gender biases without compromising performance. This approach not only enhances the fairness of VLMs applications but also preserves their efficiency and utility across diverse scenarios.

arXiv:2410.05346

AnyAttack: Towards Large-scale Self-supervised Adversarial Attacks on Vision-language Models

Due to their multimodal capabilities, Vision-Language Models (VLMs) have found numerous impactful applications in real-world scenarios. However, recent studies have revealed that VLMs are vulnerable to image-based adversarial attacks. Traditional targeted adversarial attacks require specific targets and labels, limiting their real-world impact.We present AnyAttack, a self-supervised framework that transcends the limitations of conventional attacks through a novel foundation model approach. By pre-training on the massive LAION-400M dataset without label supervision, AnyAttack achieves unprecedented flexibility - enabling any image to be transformed into an attack vector targeting any desired output across different VLMs.This approach fundamentally changes the threat landscape, making adversarial capabilities accessible at an unprecedented scale. Our extensive validation across five open-source VLMs (CLIP, BLIP, BLIP2, InstructBLIP, and MiniGPT-4) demonstrates AnyAttack's effectiveness across diverse multimodal tasks. Most concerning, AnyAttack seamlessly transfers to commercial systems including Google Gemini, Claude Sonnet, Microsoft Copilot and OpenAI GPT, revealing a systemic vulnerability requiring immediate attention.

arXiv:2410.05997공식 코드 · Papers with Code

An Eye for an Ear: Zero-shot Audio Description Leveraging an Image Captioner using Audiovisual Distribution Alignment

Multimodal large language models have fueled progress in image captioning. These models, fine-tuned on vast image datasets, exhibit a deep understanding of semantic concepts. In this work, we show that this ability can be re-purposed for audio captioning, where the joint image-language decoder can be leveraged to describe auditory content associated with image sequences within videos featuring audiovisual content. This can be achieved via multimodal alignment. Yet, this multimodal alignment task is non-trivial due to the inherent disparity between audible and visible elements in real-world videos. Moreover, multimodal representation learning often relies on contrastive learning, facing the challenge of the so-called modality gap which hinders smooth integration between modalities. In this work, we introduce a novel methodology for bridging the audiovisual modality gap by matching the distributions of tokens produced by an audio backbone and those of an image captioner. Our approach aligns the audio token distribution with that of the image tokens, enabling the model to perform zero-shot audio captioning in an unsupervised fashion while keeping the initial image captioning component unaltered. This alignment allows for the use of either audio or audiovisual input by combining or substituting the image encoder with the aligned audio encoder. Our method achieves significantly improved performances in zero-shot audio captioning, compared to existing approaches.

arXiv:2410.05800공식 코드 · Papers with Code

Core Tokensets for Data-efficient Sequential Training of Transformers

Deep networks are frequently tuned to novel tasks and continue learning from ongoing data streams. Such sequential training requires consolidation of new and past information, a challenge predominantly addressed by retaining the most important data points - formally known as coresets. Traditionally, these coresets consist of entire samples, such as images or sentences. However, recent transformer architectures operate on tokens, leading to the famous assertion that an image is worth 16x16 words. Intuitively, not all of these tokens are equally informative or memorable. Going beyond coresets, we thus propose to construct a deeper-level data summary on the level of tokens. Our respectively named core tokensets both select the most informative data points and leverage feature attribution to store only their most relevant features. We demonstrate that core tokensets yield significant performance retention in incremental image classification, open-ended visual question answering, and continual image captioning with significantly reduced memory. In fact, we empirically find that a core tokenset of 1\% of the data performs comparably to at least a twice as large and up to 10 times larger coreset.

arXiv:2410.04433공식 코드 · Papers with Code

CAPEEN: Image Captioning with Early Exits and Knowledge Distillation

Deep neural networks (DNNs) have made significant progress in recognizing visual elements and generating descriptive text in image-captioning tasks. However, their improved performance comes from increased computational burden and inference latency. Early Exit (EE) strategies can be used to enhance their efficiency, but their adaptation presents challenges in image captioning as it requires varying levels of semantic information for accurate predictions. To overcome this, we introduce CAPEEN to improve the performance of EE strategies using knowledge distillation. Inference in CAPEEN is completed at intermediary layers if prediction confidence exceeds a predefined value learned from the training data. To account for real-world deployments, where target distributions could drift from that of training samples, we introduce a variant A-CAPEEN to adapt the thresholds on the fly using Multiarmed bandits framework. Experiments on the MS COCO and Flickr30k datasets show that CAPEEN gains speedup of 1.77x while maintaining competitive performance compared to the final layer, and A-CAPEEN additionally offers robustness against distortions. The source code is available at https://github.com/Div290/CapEEN

arXiv:2410.03051

AuroraCap: Efficient, Performant Video Detailed Captioning and a New Benchmark

Video detailed captioning is a key task which aims to generate comprehensive and coherent textual descriptions of video content, benefiting both video understanding and generation. In this paper, we propose AuroraCap, a video captioner based on a large multimodal model. We follow the simplest architecture design without additional parameters for temporal modeling. To address the overhead caused by lengthy video sequences, we implement the token merging strategy, reducing the number of input visual tokens. Surprisingly, we found that this strategy results in little performance loss. AuroraCap shows superior performance on various video and image captioning benchmarks, for example, obtaining a CIDEr of 88.9 on Flickr30k, beating GPT-4V (55.3) and Gemini-1.5 Pro (82.2). However, existing video caption benchmarks only include simple descriptions, consisting of a few dozen words, which limits research in this field. Therefore, we develop VDC, a video detailed captioning benchmark with over one thousand carefully annotated structured captions. In addition, we propose a new LLM-assisted metric VDCscore for bettering evaluation, which adopts a divide-and-conquer strategy to transform long caption evaluation into multiple short question-answer pairs. With the help of human Elo ranking, our experiments show that this benchmark better correlates with human judgments of video detailed captioning quality.

arXiv:2410.02027

Quantifying the Gaps Between Translation and Native Perception in Training for Multimodal, Multilingual Retrieval

There is a scarcity of multilingual vision-language models that properly account for the perceptual differences that are reflected in image captions across languages and cultures. In this work, through a multimodal, multilingual retrieval case study, we quantify the existing lack of model flexibility. We empirically show performance gaps between training on captions that come from native German perception and captions that have been either machine-translated or human-translated from English into German. To address these gaps, we further propose and evaluate caption augmentation strategies. While we achieve mean recall improvements (+1.3), gaps still remain, indicating an open area of future work for the community.

arXiv:2410.01264

Backdooring Vision-Language Models with Out-Of-Distribution Data

The emergence of Vision-Language Models (VLMs) represents a significant advancement in integrating computer vision with Large Language Models (LLMs) to generate detailed text descriptions from visual inputs. Despite their growing importance, the security of VLMs, particularly against backdoor attacks, is under explored. Moreover, prior works often assume attackers have access to the original training data, which is often unrealistic. In this paper, we address a more practical and challenging scenario where attackers must rely solely on Out-Of-Distribution (OOD) data. We introduce VLOOD (Backdooring Vision-Language Models with Out-of-Distribution Data), a novel approach with two key contributions: (1) demonstrating backdoor attacks on VLMs in complex image-to-text tasks while minimizing degradation of the original semantics under poisoned inputs, and (2) proposing innovative techniques for backdoor injection without requiring any access to the original training data. Our evaluation on image captioning and visual question answering (VQA) tasks confirms the effectiveness of VLOOD, revealing a critical security vulnerability in VLMs and laying the foundation for future research on securing multimodal models against sophisticated threats.

arXiv:2409.19232

TrojVLM: Backdoor Attack Against Vision Language Models

The emergence of Vision Language Models (VLMs) is a significant advancement in integrating computer vision with Large Language Models (LLMs) to produce detailed text descriptions based on visual inputs, yet it introduces new security vulnerabilities. Unlike prior work that centered on single modalities or classification tasks, this study introduces TrojVLM, the first exploration of backdoor attacks aimed at VLMs engaged in complex image-to-text generation. Specifically, TrojVLM inserts predetermined target text into output text when encountering poisoned images. Moreover, a novel semantic preserving loss is proposed to ensure the semantic integrity of the original image content. Our evaluation on image captioning and visual question answering (VQA) tasks confirms the effectiveness of TrojVLM in maintaining original semantic content while triggering specific target text outputs. This study not only uncovers a critical security risk in VLMs and image-to-text generation but also sets a foundation for future research on securing multimodal models against such sophisticated threats.

arXiv:2409.19960공식 코드 · Papers with Code

TROPE: TRaining-Free Object-Part Enhancement for Seamlessly Improving Fine-Grained Zero-Shot Image Captioning

Zero-shot inference, where pre-trained models perform tasks without specific training data, is an exciting emergent ability of large models like CLIP. Although there has been considerable exploration into enhancing zero-shot abilities in image captioning (IC) for popular datasets such as MSCOCO and Flickr8k, these approaches fall short with fine-grained datasets like CUB, FLO, UCM-Captions, and Sydney-Captions. These datasets require captions to discern between visually and semantically similar classes, focusing on detailed object parts and their attributes. To overcome this challenge, we introduce TRaining-Free Object-Part Enhancement (TROPE). TROPE enriches a base caption with additional object-part details using object detector proposals and Natural Language Processing techniques. It complements rather than alters the base caption, allowing seamless integration with other captioning methods and offering users enhanced flexibility. Our evaluations show that TROPE consistently boosts performance across all tested zero-shot IC approaches and achieves state-of-the-art results on fine-grained IC datasets.

arXiv:2409.19255

DENEB: A Hallucination-Robust Automatic Evaluation Metric for Image Captioning

In this work, we address the challenge of developing automatic evaluation metrics for image captioning, with a particular focus on robustness against hallucinations. Existing metrics are often inadequate for handling hallucinations, primarily due to their limited ability to compare candidate captions with multifaceted reference captions. To address this shortcoming, we propose DENEB, a novel supervised automatic evaluation metric specifically robust against hallucinations. DENEB incorporates the Sim-Vec Transformer, a mechanism that processes multiple references simultaneously, thereby efficiently capturing the similarity between an image, a candidate caption, and reference captions. To train DENEB, we construct the diverse and balanced Nebula dataset comprising 32,978 images, paired with human judgments provided by 805 annotators. We demonstrated that DENEB achieves state-of-the-art performance among existing LLM-free metrics on the FOIL, Composite, Flickr8K-Expert, Flickr8K-CF, Nebula, and PASCAL-50S datasets, validating its effectiveness and robustness against hallucinations.

arXiv:2409.18753

Enhancing Explainability in Multimodal Large Language Models Using Ontological Context

Recently, there has been a growing interest in Multimodal Large Language Models (MLLMs) due to their remarkable potential in various tasks integrating different modalities, such as image and text, as well as applications such as image captioning and visual question answering. However, such models still face challenges in accurately captioning and interpreting specific visual concepts and classes, particularly in domain-specific applications. We argue that integrating domain knowledge in the form of an ontology can significantly address these issues. In this work, as a proof of concept, we propose a new framework that combines ontology with MLLMs to classify images of plant diseases. Our method uses concepts about plant diseases from an existing disease ontology to query MLLMs and extract relevant visual concepts from images. Then, we use the reasoning capabilities of the ontology to classify the disease according to the identified concepts. Ensuring that the model accurately uses the concepts describing the disease is crucial in domain-specific applications. By employing an ontology, we can assist in verifying this alignment. Additionally, using the ontology's inference capabilities increases transparency, explainability, and trust in the decision-making process while serving as a judge by checking if the annotations of the concepts by MLLMs are aligned with those in the ontology and displaying the rationales behind their errors. Our framework offers a new direction for synergizing ontologies and MLLMs, supported by an empirical study using different well-known MLLMs.

arXiv:2409.18467

A TextGCN-Based Decoding Approach for Improving Remote Sensing Image Captioning

Remote sensing images are highly valued for their ability to address complex real-world issues such as risk management, security, and meteorology. However, manually captioning these images is challenging and requires specialized knowledge across various domains. This letter presents an approach for automatically describing (captioning) remote sensing images. We propose a novel encoder-decoder setup that deploys a Text Graph Convolutional Network (TextGCN) and multi-layer LSTMs. The embeddings generated by TextGCN enhance the decoder's understanding by capturing the semantic relationships among words at both the sentence and corpus levels. Furthermore, we advance our approach with a comparison-based beam search method to ensure fairness in the search strategy for generating the final caption. We present an extensive evaluation of our approach against various other state-of-the-art encoder-decoder frameworks. We evaluated our method across three datasets using seven metrics: BLEU-1 to BLEU-4, METEOR, ROUGE-L, and CIDEr. The results demonstrate that our approach significantly outperforms other state-of-the-art encoder-decoder methods.

arXiv:2409.18046공식 코드 · Papers with Code

IFCap: Image-like Retrieval and Frequency-based Entity Filtering for Zero-shot Captioning

Recent advancements in image captioning have explored text-only training methods to overcome the limitations of paired image-text data. However, existing text-only training methods often overlook the modality gap between using text data during training and employing images during inference. To address this issue, we propose a novel approach called Image-like Retrieval, which aligns text features with visually relevant features to mitigate the modality gap. Our method further enhances the accuracy of generated captions by designing a Fusion Module that integrates retrieved captions with input features. Additionally, we introduce a Frequency-based Entity Filtering technique that significantly improves caption quality. We integrate these methods into a unified framework, which we refer to as IFCap ($\textbf{I}$mage-like Retrieval and $\textbf{F}$requency-based Entity Filtering for Zero-shot $\textbf{Cap}$tioning). Through extensive experimentation, our straightforward yet powerful approach has demonstrated its efficacy, outperforming the state-of-the-art methods by a significant margin in both image captioning and video captioning compared to zero-shot captioning based on text-only training.

arXiv:2409.17146공식 코드 · Papers with Code

Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models

Today's most advanced vision-language models (VLMs) remain proprietary. The strongest open-weight models rely heavily on synthetic data from proprietary VLMs to achieve good performance, effectively distilling these closed VLMs into open ones. As a result, the community has been missing foundational knowledge about how to build performant VLMs from scratch. We present Molmo, a new family of VLMs that are state-of-the-art in their class of openness. Our key contribution is a collection of new datasets called PixMo, including a dataset of highly detailed image captions for pre-training, a free-form image Q&A dataset for fine-tuning, and an innovative 2D pointing dataset, all collected without the use of external VLMs. The success of our approach relies on careful modeling choices, a well-tuned training pipeline, and, most critically, the quality of our newly collected datasets. Our best-in-class 72B model not only outperforms others in the class of open weight and data models, but also outperforms larger proprietary models including Claude 3.5 Sonnet, and Gemini 1.5 Pro and Flash, second only to GPT-4o based on both academic benchmarks and on a large human evaluation. Our model weights, new datasets, and source code are available at https://molmo.allenai.org/blog.