In the real world, documents are organized in different formats and varied modalities. Traditional retrieval pipelines require tailored document parsing techniques and content extraction modules to prepare input for indexing. This process is tedious, prone to errors, and has information loss. To this end, we propose Document Screenshot Embedding (DSE), a novel retrieval paradigm that regards document screenshots as a unified input format, which does not require any content extraction preprocess and preserves all the information in a document (e.g., text, image and layout). DSE leverages a large vision-language model to directly encode document screenshots into dense representations for retrieval. To evaluate our method, we first craft the dataset of Wiki-SS, a 1.3M Wikipedia web page screenshots as the corpus to answer the questions from the Natural Questions dataset. In such a text-intensive document retrieval setting, DSE shows competitive effectiveness compared to other text retrieval methods relying on parsing. For example, DSE outperforms BM25 by 17 points in top-1 retrieval accuracy. Additionally, in a mixed-modality task of slide retrieval, DSE significantly outperforms OCR text retrieval methods by over 15 points in nDCG@10. These experiments show that DSE is an effective document retrieval paradigm for diverse types of documents. Model checkpoints, code, and Wiki-SS collection will be released.
We study the visual semantic embedding problem for image-text matching. Most existing work utilizes a tailored cross-attention mechanism to perform local alignment across the two image and text modalities. This is computationally expensive, even though it is more powerful than the unimodal dual-encoder approach. This work introduces a dual-encoder image-text matching model, leveraging a scene graph to represent captions with nodes for objects and attributes interconnected by relational edges. Utilizing a graph attention network, our model efficiently encodes object-attribute and object-object semantic relations, resulting in a robust and fast-performing system. Representing caption as a scene graph offers the ability to utilize the strong relational inductive bias of graph neural networks to learn object-attribute and object-object relations effectively. To train the model, we propose losses that align the image and caption both at the holistic level (image-caption) and the local level (image-object entity), which we show is key to the success of the model. Our model is termed Composition model for Object Relations and Attributes, CORA. Experimental results on two prominent image-text retrieval benchmarks, Flickr30K and MSCOCO, demonstrate that CORA outperforms existing state-of-the-art computationally expensive cross-attention methods regarding recall score while achieving fast computation speed of the dual encoder.
Retrieving and extracting knowledge from extensive research documents and large databases presents significant challenges for researchers, students, and professionals in today's information-rich era. Existing retrieval systems, which rely on general-purpose Large Language Models (LLMs), often fail to provide accurate responses to domain-specific inquiries. Additionally, the high cost of pretraining or fine-tuning LLMs for specific domains limits their widespread adoption. To address these limitations, we propose a novel methodology that combines the generative capabilities of LLMs with the fast and accurate retrieval capabilities of vector databases. This advanced retrieval system can efficiently handle both tabular and non-tabular data, understand natural language user queries, and retrieve relevant information without fine-tuning. The developed model, Generative Text Retrieval (GTR), is adaptable to both unstructured and structured data with minor refinement. GTR was evaluated on both manually annotated and public datasets, achieving over 90% accuracy and delivering truthful outputs in 87% of cases. Our model achieved state-of-the-art performance with a Rouge-L F1 score of 0.98 on the MSMARCO dataset. The refined model, Generative Tabular Text Retrieval (GTR-T), demonstrated its efficiency in large database querying, achieving an Execution Accuracy (EX) of 0.82 and an Exact-Set-Match (EM) accuracy of 0.60 on the Spider dataset, using an open-source LLM. These efforts leverage Generative AI and In-Context Learning to enhance human-text interaction and make advanced AI capabilities more accessible. By integrating robust retrieval systems with powerful LLMs, our approach aims to democratize access to sophisticated AI tools, improving the efficiency, accuracy, and scalability of AI-driven information retrieval and database querying.
Existing Vision-Language Compositionality (VLC) benchmarks like SugarCrepe are formulated as image-to-text retrieval problems, where, given an image, the models need to select between the correct textual description and a synthetic hard negative text. In this work, we present the Bidirectional Vision-Language Compositionality (BiVLC) dataset. The novelty of BiVLC is to add a synthetic hard negative image generated from the synthetic text, resulting in two image-to-text retrieval examples (one for each image) and, more importantly, two text-to-image retrieval examples (one for each text). Human annotators filter out ill-formed examples ensuring the validity of the benchmark. The experiments on BiVLC uncover a weakness of current multimodal models, as they perform poorly in the text-to-image direction. In fact, when considering both retrieval directions, the conclusions obtained in previous works change significantly. In addition to the benchmark, we show that a contrastive model trained using synthetic images and texts significantly improves over the base model in SugarCrepe and in BiVLC for both retrieval directions. The gap to human performance in BiVLC confirms that Vision-Language Compositionality is still a challenging problem. BiVLC and code are available at https://imirandam.github.io/BiVLC_project_page.
Vision-Language Foundation Models (VLFMs) have made remarkable progress on various multimodal tasks, such as image captioning, image-text retrieval, visual question answering, and visual grounding. However, most methods rely on training with general image datasets, and the lack of geospatial data leads to poor performance on earth observation. Numerous geospatial image-text pair datasets and VLFMs fine-tuned on them have been proposed recently. These new approaches aim to leverage large-scale, multimodal geospatial data to build versatile intelligent models with diverse geo-perceptive capabilities, which we refer to as Vision-Language Geo-Foundation Models (VLGFMs). This paper thoroughly reviews VLGFMs, summarizing and analyzing recent developments in the field. In particular, we introduce the background and motivation behind the rise of VLGFMs, highlighting their unique research significance. Then, we systematically summarize the core technologies employed in VLGFMs, including data construction, model architectures, and applications of various multimodal geospatial tasks. Finally, we conclude with insights, issues, and discussions regarding future research directions. To the best of our knowledge, this is the first comprehensive literature review of VLGFMs. We keep tracing related works at https://github.com/zytx121/Awesome-VLGFM.
Audio-text retrieval is a challenging task, requiring the search for an audio clip or a text caption within a database. The predominant focus of existing research on English descriptions poses a limitation on the applicability of such models, given the abundance of non-English content in real-world data. To address these linguistic disparities, we propose a language enhancement (LE), using a multilingual text encoder (SONAR) to encode the text data with language-specific information. Additionally, we optimize the audio encoder through the application of consistent ensemble distillation (CED), enhancing support for variable-length audio-text retrieval. Our methodology excels in English audio-text retrieval, demonstrating state-of-the-art (SOTA) performance on commonly used datasets such as AudioCaps and Clotho. Simultaneously, the approach exhibits proficiency in retrieving content in seven other languages with only 10% of additional language-enhanced training data, yielding promising results. The source code is publicly available https://github.com/zyyan4/ml-clap.
Contrastive Language-Image Pre-training (CLIP) has significantly improved performance in various vision-language tasks by expanding the dataset with image-text pairs obtained from websites. This paper further explores CLIP from the perspectives of data and model architecture. To address the prevalence of noisy data and enhance the quality of large-scale image-text data crawled from the internet, we introduce a diverse description generation framework that can leverage Large Language Models (LLMs) to synthesize and refine content from web-based texts, synthetic captions, and detection tags. Furthermore, we propose RWKV-CLIP, the first RWKV-driven vision-language representation learning model that combines the effective parallel training of transformers with the efficient inference of RNNs. Comprehensive experiments across various model scales and pre-training datasets demonstrate that RWKV-CLIP is a robust and efficient vision-language representation learner, it achieves state-of-the-art performance in several downstream tasks, including linear probe, zero-shot classification, and zero-shot image-text retrieval. To facilitate future research, the code and pre-trained models are released at https://github.com/deepglint/RWKV-CLIP
In this demonstration, we present Country Guesser, a live system that guesses the country that a photo is taken in. In particular, given a Google Street View image, our federated ranking model uses a combination of computer vision, machine learning and text retrieval methods to compute a ranking of likely countries of the location shown in a given image from Street View. Interestingly, using text-based features to probe large pre-trained language models can assist to provide cross-modal supervision. We are not aware of previous country guessing systems informed by visual and textual features.
Text-based person retrieval (TPR) is a challenging task that involves retrieving a specific individual based on a textual description. Despite considerable efforts to bridge the gap between vision and language, the significant differences between these modalities continue to pose a challenge. Previous methods have attempted to align text and image samples in a modal-shared space, but they face uncertainties in optimization directions due to the movable features of both modalities and the failure to account for one-to-many relationships of image-text pairs in TPR datasets. To address this issue, we propose an effective bi-directional one-to-many embedding paradigm that offers a clear optimization direction for each sample, thus mitigating the optimization problem. Additionally, this embedding scheme generates multiple features for each sample without introducing trainable parameters, making it easier to align with several positive samples. Based on this paradigm, we propose a novel Bi-directional one-to-many Embedding Alignment (Beat) model to address the TPR task. Our experimental results demonstrate that the proposed Beat model achieves state-of-the-art performance on three popular TPR datasets, including CUHK-PEDES (65.61 R@1), ICFG-PEDES (58.25 R@1), and RSTPReID (48.10 R@1). Furthermore, additional experiments on MS-COCO, CUB, and Flowers datasets further demonstrate the potential of Beat to be applied to other image-text retrieval tasks.
Videos are more informative than images because they capture the dynamics of the scene. By representing motion in videos, we can capture dynamic activities. In this work, we introduce GPT-4 generated motion descriptions that capture fine-grained motion descriptions of activities and apply them to three action datasets. We evaluated several video-text models on the task of retrieval of motion descriptions. We found that they fall far behind human expert performance on two action datasets, raising the question of whether video-text models understand motion in videos. To address it, we introduce a method of improving motion understanding in video-text models by utilizing motion descriptions. This method proves to be effective on two action datasets for the motion description retrieval task. The results draw attention to the need for quality captions involving fine-grained motion information in existing datasets and demonstrate the effectiveness of the proposed pipeline in understanding fine-grained motion during video-text retrieval.
We propose a new "bi-metric" framework for designing nearest neighbor data structures. Our framework assumes two dissimilarity functions: a ground-truth metric that is accurate but expensive to compute, and a proxy metric that is cheaper but less accurate. In both theory and practice, we show how to construct data structures using only the proxy metric such that the query procedure achieves the accuracy of the expensive metric, while only using a limited number of calls to both metrics. Our theoretical results instantiate this framework for two popular nearest neighbor search algorithms: DiskANN and Cover Tree. In both cases we show that, as long as the proxy metric used to construct the data structure approximates the ground-truth metric up to a bounded factor, our data structure achieves arbitrarily good approximation guarantees with respect to the ground-truth metric. On the empirical side, we apply the framework to the text retrieval problem with two dissimilarity functions evaluated by ML models with vastly different computational costs. We observe that for almost all data sets in the MTEB benchmark, our approach achieves a considerably better accuracy-efficiency tradeoff than the alternatives, such as re-ranking.
In this paper, we study a challenging but less-touched problem in cross-modal retrieval, i.e., partially mismatched pairs (PMPs). Specifically, in real-world scenarios, a huge number of multimedia data (e.g., the Conceptual Captions dataset) are collected from the Internet, and thus it is inevitable to wrongly treat some irrelevant cross-modal pairs as matched. Undoubtedly, such a PMP problem will remarkably degrade the cross-modal retrieval performance. To tackle this problem, we derive a unified theoretical Robust Cross-modal Learning framework (RCL) with an unbiased estimator of the cross-modal retrieval risk, which aims to endow the cross-modal retrieval methods with robustness against PMPs. In detail, our RCL adopts a novel complementary contrastive learning paradigm to address the following two challenges, i.e., the overfitting and underfitting issues. On the one hand, our method only utilizes the negative information which is much less likely false compared with the positive information, thus avoiding the overfitting issue to PMPs. However, these robust strategies could induce underfitting issues, thus making training models more difficult. On the other hand, to address the underfitting issue brought by weak supervision, we present to leverage of all available negative pairs to enhance the supervision contained in the negative information. Moreover, to further improve the performance, we propose to minimize the upper bounds of the risk to pay more attention to hard samples. To verify the effectiveness and robustness of the proposed method, we carry out comprehensive experiments on five widely-used benchmark datasets compared with nine state-of-the-art approaches w.r.t. the image-text and video-text retrieval tasks. The code is available at https://github.com/penghu-cs/RCL .
Current video-language models (VLMs) rely extensively on instance-level alignment between video and language modalities, which presents two major limitations: (1) visual reasoning disobeys the natural perception that humans do in first-person perspective, leading to a lack of reasoning interpretation; and (2) learning is limited in capturing inherent fine-grained relationships between two modalities. In this paper, we take an inspiration from human perception and explore a compositional approach for egocentric video representation. We introduce HENASY (Hierarchical ENtities ASsemblY), which includes a spatiotemporal token grouping mechanism to explicitly assemble dynamically evolving scene entities through time and model their relationship for video representation. By leveraging compositional structure understanding, HENASY possesses strong interpretability via visual grounding with free-form text queries. We further explore a suite of multi-grained contrastive losses to facilitate entity-centric understandings. This comprises three alignment types: video-narration, noun-entity, verb-entities alignments. Our method demonstrates strong interpretability in both quantitative and qualitative experiments; while maintaining competitive performances on five downstream tasks via zero-shot transfer or as video/text representation, including video/text retrieval, action recognition, multi-choice query, natural language query, and moments query.
CLIP4Clip model transferred from the CLIP has been the de-factor standard to solve the video clip retrieval task from frame-level input, triggering the surge of CLIP4Clip-based models in the video-text retrieval domain. In this work, we rethink the inherent limitation of widely-used mean pooling operation in the frame features aggregation and investigate the adaptions of excitation and aggregation design for discriminative video representation generation. We present a novel excitationand-aggregation design, including (1) The excitation module is available for capturing non-mutuallyexclusive relationships among frame features and achieving frame-wise features recalibration, and (2) The aggregation module is applied to learn exclusiveness used for frame representations aggregation. Similarly, we employ the cascade of sequential module and aggregation design to generate discriminative video representation in the sequential type. Besides, we adopt the excitation design in the tight type to obtain representative frame features for multi-modal interaction. The proposed modules are evaluated on three benchmark datasets of MSR-VTT, ActivityNet and DiDeMo, achieving MSR-VTT (43.9 R@1), ActivityNet (44.1 R@1) and DiDeMo (31.0 R@1). They outperform the CLIP4Clip results by +1.2% (+0.5%), +4.5% (+1.9%) and +9.5% (+2.7%) relative (absolute) improvements, demonstrating the superiority of our proposed excitation and aggregation designs. We hope our work will serve as an alternative for frame representations aggregation and facilitate future research.
Sign language video retrieval plays a key role in facilitating information access for the deaf community. Despite significant advances in video-text retrieval, the complexity and inherent uncertainty of sign language preclude the direct application of these techniques. Previous methods achieve the mapping between sign language video and text through fine-grained modal alignment. However, due to the scarcity of fine-grained annotation, the uncertainty inherent in sign language video is underestimated, limiting the further development of sign language retrieval tasks. To address this challenge, we propose a novel Uncertainty-aware Probability Distribution Retrieval (UPRet), that conceptualizes the mapping process of sign language video and text in terms of probability distributions, explores their potential interrelationships, and enables flexible mappings. Experiments on three benchmarks demonstrate the effectiveness of our method, which achieves state-of-the-art results on How2Sign (59.1%), PHOENIX-2014T (72.0%), and CSL-Daily (78.4%).
A visual-language model (VLM) pre-trained on natural images and text pairs poses a significant barrier when applied to medical contexts due to domain shift. Yet, adapting or fine-tuning these VLMs for medical use presents considerable hurdles, including domain misalignment, limited access to extensive datasets, and high-class imbalances. Hence, there is a pressing need for strategies to effectively adapt these VLMs to the medical domain, as such adaptations would prove immensely valuable in healthcare applications. In this study, we propose a framework designed to adeptly tailor VLMs to the medical domain, employing selective sampling and hard-negative mining techniques for enhanced performance in retrieval tasks. We validate the efficacy of our proposed approach by implementing it across two distinct VLMs: the in-domain VLM (MedCLIP) and out-of-domain VLMs (ALBEF). We assess the performance of these models both in their original off-the-shelf state and after undergoing our proposed training strategies, using two extensive datasets containing mammograms and their corresponding reports. Our evaluation spans zero-shot, few-shot, and supervised scenarios. Through our approach, we observe a notable enhancement in Recall@K performance for the image-text retrieval task.
Contrastive Language-Image Pretraining (CLIP) is widely used to train models to align images and texts in a common embedding space by mapping them to fixed-sized vectors. These models are key to multimodal information retrieval and related tasks. However, CLIP models generally underperform in text-only tasks compared to specialized text models. This creates inefficiencies for information retrieval systems that keep separate embeddings and models for text-only and multimodal tasks. We propose a novel, multi-task contrastive training method to address this issue, which we use to train the jina-clip-v1 model to achieve the state-of-the-art performance on both text-image and text-text retrieval tasks.
Remote Sensing Image-Text Retrieval (RSITR) is pivotal for knowledge services and data mining in the remote sensing (RS) domain. Considering the multi-scale representations in image content and text vocabulary can enable the models to learn richer representations and enhance retrieval. Current multi-scale RSITR approaches typically align multi-scale fused image features with text features, but overlook aligning image-text pairs at distinct scales separately. This oversight restricts their ability to learn joint representations suitable for effective retrieval. We introduce a novel Multi-Scale Alignment (MSA) method to overcome this limitation. Our method comprises three key innovations: (1) Multi-scale Cross-Modal Alignment Transformer (MSCMAT), which computes cross-attention between single-scale image features and localized text features, integrating global textual context to derive a matching score matrix within a mini-batch, (2) a multi-scale cross-modal semantic alignment loss that enforces semantic alignment across scales, and (3) a cross-scale multi-modal semantic consistency loss that uses the matching matrix from the largest scale to guide alignment at smaller scales. We evaluated our method across multiple datasets, demonstrating its efficacy with various visual backbones and establishing its superiority over existing state-of-the-art methods. The GitHub URL for our project is: https://github.com/yr666666/MSA
Pre-trained vision-language (VL) models are highly vulnerable to adversarial attacks. However, existing defense methods primarily focus on image classification, overlooking two key aspects of VL tasks: multimodal attacks, where both image and text can be perturbed, and the one-to-many relationship of images and texts, where a single image can correspond to multiple textual descriptions and vice versa (1:N and N:1). This work is the first to explore defense strategies against multimodal attacks in VL tasks, whereas prior VL defense methods focus on vision robustness. We propose multimodal adversarial training (MAT), which incorporates adversarial perturbations in both image and text modalities during training, significantly outperforming existing unimodal defenses. Furthermore, we discover that MAT is limited by deterministic one-to-one (1:1) image-text pairs in VL training data. To address this, we conduct a comprehensive study on leveraging one-to-many relationships to enhance robustness, investigating diverse augmentation techniques. Our analysis shows that, for a more effective defense, augmented image-text pairs should be well-aligned, diverse, yet avoid distribution shift -- conditions overlooked by prior research. Our experiments show that MAT can effectively be applied to different VL models and tasks to improve adversarial robustness, outperforming previous efforts. Our code will be made public upon acceptance.
With the emergence of diffusion models as the frontline of generative models, many researchers have proposed molecule generation techniques with conditional diffusion models. However, the unavoidable discreteness of a molecule makes it difficult for a diffusion model to connect raw data with highly complex conditions like natural language. To address this, we present a novel latent diffusion model dubbed LDMol for text-conditioned molecule generation. LDMol comprises a molecule autoencoder that produces a learnable and structurally informative feature space, and a natural language-conditioned latent diffusion model. In particular, recognizing that multiple SMILES notations can represent the same molecule, we employ a contrastive learning strategy to extract feature space that is aware of the unique characteristics of the molecule structure. LDMol outperforms the existing baselines on the text-to-molecule generation benchmark, suggesting a potential for diffusion models can outperform autoregressive models in text data generation with a better choice of the latent domain. Furthermore, we show that LDMol can be applied to downstream tasks such as molecule-to-text retrieval and text-guided molecule editing, demonstrating its versatility as a diffusion model.