Recently, the remarkable success of pre-trained Vision Transformers (ViTs) from image-text matching has sparked an interest in image-to-video adaptation. However, most current approaches retain the full forward pass for each frame, leading to a high computation overhead for processing entire videos. In this paper, we present InTI, a novel approach for compressive image-to-video adaptation using dynamic Inter-frame Token Interpolation. InTI aims to softly preserve the informative tokens without disrupting their coherent spatiotemporal structure. Specifically, each token pair at identical positions within neighbor frames is linearly aggregated into a new token, where the aggregation weights are generated by a multi-scale context-aware network. In this way, the information of neighbor frames can be adaptively compressed in a point-by-point manner, thereby effectively reducing the number of processed frames by half each time. Importantly, InTI can be seamlessly integrated with existing adaptation methods, achieving strong performance without extra-complex design. On Kinetics-400, InTI reaches a top-1 accuracy of 87.1 with a remarkable 37.5% reduction in GFLOPs compared to naive adaptation. When combined with additional temporal modules, InTI achieves a top-1 accuracy of 87.6 with a 37% reduction in GFLOPs. Similar conclusions have been verified in other common datasets.
Ordinal regression is a fundamental problem within the field of computer vision, with customised well-trained models on specific tasks. While pre-trained vision-language models (VLMs) have exhibited impressive performance on various vision tasks, their potential for ordinal regression has received less exploration. In this study, we first investigate CLIP's potential for ordinal regression, from which we expect the model could generalise to different ordinal regression tasks and scenarios. Unfortunately, vanilla CLIP fails on this task, since current VLMs have a well-documented limitation of encapsulating compositional concepts such as number sense. We propose a simple yet effective method called NumCLIP to improve the quantitative understanding of VLMs. We disassemble the exact image to number-specific text matching problem into coarse classification and fine prediction stages. We discretize and phrase each numerical bin with common language concept to better leverage the available pre-trained alignment in CLIP. To consider the inherent continuous property of ordinal regression, we propose a novel fine-grained cross-modal ranking-based regularisation loss specifically designed to keep both semantic and ordinal alignment in CLIP's feature space. Experimental results on three general ordinal regression tasks demonstrate the effectiveness of NumCLIP, with 10% and 3.83% accuracy improvement on historical image dating and image aesthetics assessment task, respectively. Code is publicly available at https://github.com/xmed-lab/NumCLIP.
We address the problem of detecting and mapping all books in a collection of images to entries in a given book catalogue. Instead of performing independent retrieval for each book detected, we treat the image-text mapping problem as a many-to-many matching process, looking for the best overall match between the two sets. We combine a state-of-the-art segmentation method (SAM) to detect book spines and extract book information using a commercial OCR. We then propose a two-stage approach for text-image matching, where CLIP embeddings are used first for fast matching, followed by a second slower stage to refine the matching, employing either the Hungarian Algorithm or a BERT-based model trained to cope with noisy OCR input and partial text matches. To evaluate our approach, we publish a new dataset of annotated bookshelf images that covers the whole book collection of a public library in Spain. In addition, we provide two target lists of book metadata, a closed-set of 15k book titles that corresponds to the known library inventory, and an open-set of 2.3M book titles to simulate an open-world scenario. We report results on two settings, on one hand on a matching-only task, where the book segments and OCR is given and the objective is to perform many-to-many matching against the target lists, and a combined detection and matching task, where books must be first detected and recognised before they are matched to the target list entries. We show that both the Hungarian Matching and the proposed BERT-based model outperform a fuzzy string matching baseline, and we highlight inherent limitations of the matching algorithms as the target increases in size, and when either of the two sets (detected books or target book list) is incomplete. The dataset and code are available at https://github.com/llabres/library-dataset
In dense retrieval, embedding long texts into dense vectors can result in information loss, leading to inaccurate query-text matching. Additionally, low-quality texts with excessive noise or sparse key information are unlikely to align well with relevant queries. Recent studies mainly focus on improving the sentence embedding model or retrieval process. In this work, we introduce a novel text augmentation framework for dense retrieval. This framework transforms raw documents into information-dense text formats, which supplement the original texts to effectively address the aforementioned issues without modifying embedding or retrieval methodologies. Two text representations are generated via large language models (LLMs) zero-shot prompting: question-answer pairs and element-driven events. We term this approach QAEA-DR: unifying question-answer generation and event extraction in a text augmentation framework for dense retrieval. To further enhance the quality of generated texts, a scoring-based evaluation and regeneration mechanism is introduced in LLM prompting. Our QAEA-DR model has a positive impact on dense retrieval, supported by both theoretical analysis and empirical experiments.
Text relevance or text matching of query and product is an essential technique for the e-commerce search system to ensure that the displayed products can match the intent of the query. Many studies focus on improving the performance of the relevance model in search system. Recently, pre-trained language models like BERT have achieved promising performance on the text relevance task. While these models perform well on the offline test dataset, there are still obstacles to deploy the pre-trained language model to the online system as their high latency. The two-tower model is extensively employed in industrial scenarios, owing to its ability to harmonize performance with computational efficiency. Regrettably, such models present an opaque ``black box'' nature, which prevents developers from making special optimizations. In this paper, we raise deep Bag-of-Words (DeepBoW) model, an efficient and interpretable relevance architecture for Chinese e-commerce. Our approach proposes to encode the query and the product into the sparse BoW representation, which is a set of word-weight pairs. The weight means the important or the relevant score between the corresponding word and the raw text. The relevance score is measured by the accumulation of the matched word between the sparse BoW representation of the query and the product. Compared to popular dense distributed representation that usually suffers from the drawback of black-box, the most advantage of the proposed representation model is highly explainable and interventionable, which is a superior advantage to the deployment and operation of online search engines. Moreover, the online efficiency of the proposed model is even better than the most efficient inner product form of dense representation ...
Documents are visually rich structures that convey information through text, but also figures, page layouts, tables, or even fonts. Since modern retrieval systems mainly rely on the textual information they extract from document pages to index documents -often through lengthy and brittle processes-, they struggle to exploit key visual cues efficiently. This limits their capabilities in many practical document retrieval applications such as Retrieval Augmented Generation (RAG). To benchmark current systems on visually rich document retrieval, we introduce the Visual Document Retrieval Benchmark ViDoRe, composed of various page-level retrieval tasks spanning multiple domains, languages, and practical settings. The inherent complexity and performance shortcomings of modern systems motivate a new concept; doing document retrieval by directly embedding the images of the document pages. We release ColPali, a Vision Language Model trained to produce high-quality multi-vector embeddings from images of document pages. Combined with a late interaction matching mechanism, ColPali largely outperforms modern document retrieval pipelines while being drastically simpler, faster and end-to-end trainable. We release models, data, code and benchmarks under open licenses at https://hf.co/vidore.
Image-text matching (ITM) is a fundamental problem in computer vision. The key issue lies in jointly learning the visual and textual representation to estimate their similarity accurately. Most existing methods focus on feature enhancement within modality or feature interaction across modalities, which, however, neglects the contextual information of the object representation based on the inter-object relationships that match the corresponding sentences with rich contextual semantics. In this paper, we propose a Hybrid-modal Interaction with multiple Relational Enhancements (termed \textit{Hire}) for image-text matching, which correlates the intra- and inter-modal semantics between objects and words with implicit and explicit relationship modelling. In particular, the explicit intra-modal spatial-semantic graph-based reasoning network is designed to improve the contextual representation of visual objects with salient spatial and semantic relational connectivities, guided by the explicit relationships of the objects' spatial positions and their scene graph. We use implicit relationship modelling for potential relationship interactions before explicit modelling to improve the fault tolerance of explicit relationship detection. Then the visual and textual semantic representations are refined jointly via inter-modal interactive attention and cross-modal alignment. To correlate the context of objects with the textual context, we further refine the visual semantic representation via cross-level object-sentence and word-image-based interactive attention. Extensive experiments validate that the proposed hybrid-modal interaction with implicit and explicit modelling is more beneficial for image-text matching. And the proposed \textit{Hire} obtains new state-of-the-art results on MS-COCO and Flickr30K benchmarks.
Large language models (LLMs) have increased interest in vision language models (VLMs), which process image-text pairs as input. Studies investigating the visual understanding ability of VLMs have been proposed, but such studies are still preliminary because existing datasets do not permit a comprehensive evaluation of the fine-grained visual linguistic abilities of VLMs across multiple languages. To further explore the strengths of VLMs, such as GPT-4V \cite{openai2023GPT4}, we developed new datasets for the systematic and qualitative analysis of VLMs. Our contribution is four-fold: 1) we introduced nine vision-and-language (VL) tasks (including object recognition, image-text matching, and more) and constructed multilingual visual-text datasets in four languages: English, Japanese, Swahili, and Urdu through utilizing templates containing \textit{questions} and prompting GPT4-V to generate the \textit{answers} and the \textit{rationales}, 2) introduced a new VL task named \textit{unrelatedness}, 3) introduced rationales to enable human understanding of the VLM reasoning process, and 4) employed human evaluation to measure the suitability of proposed datasets for VL tasks. We show that VLMs can be fine-tuned on our datasets. Our work is the first to conduct such analyses in Swahili and Urdu. Also, it introduces \textit{rationales} in VL analysis, which played a vital role in the evaluation.
Image-text matching is a key multimodal task that aims to model the semantic association between images and text as a matching relationship. With the advent of the multimedia information age, image, and text data show explosive growth, and how to accurately realize the efficient and accurate semantic correspondence between them has become the core issue of common concern in academia and industry. In this study, we delve into the limitations of current multimodal deep learning models in processing image-text pairing tasks. Therefore, we innovatively design an advanced multimodal deep learning architecture, which combines the high-level abstract representation ability of deep neural networks for visual information with the advantages of natural language processing models for text semantic understanding. By introducing a novel cross-modal attention mechanism and hierarchical feature fusion strategy, the model achieves deep fusion and two-way interaction between image and text feature space. In addition, we also optimize the training objectives and loss functions to ensure that the model can better map the potential association structure between images and text during the learning process. Experiments show that compared with existing image-text matching models, the optimized new model has significantly improved performance on a series of benchmark data sets. In addition, the new model also shows excellent generalization and robustness on large and diverse open scenario datasets and can maintain high matching performance even in the face of previously unseen complex situations.
Pre-trained vision-language models like CLIP have shown powerful zero-shot inference ability via image-text matching and prove to be strong few-shot learners in various downstream tasks. However, in real-world scenarios, adapting CLIP to downstream tasks may encounter the following challenges: 1) data may exhibit long-tailed data distributions and might not have abundant samples for all the classes; 2) There might be emerging tasks with new classes that contain no samples at all. To overcome them, we propose a novel framework to achieve efficient and long-tailed generalization, which can be termed as Candle. During the training process, we propose compensating logit-adjusted loss to encourage large margins of prototypes and alleviate imbalance both within the base classes and between the base and new classes. For efficient adaptation, we treat the CLIP model as a black box and leverage the extracted features to obtain visual and textual prototypes for prediction. To make full use of multi-modal information, we also propose cross-modal attention to enrich the features from both modalities. For effective generalization, we introduce virtual prototypes for new classes to make up for their lack of training images. Candle achieves state-of-the-art performance over extensive experiments on 11 diverse datasets while substantially reducing the training time, demonstrating the superiority of our approach. The source code is available at https://github.com/shijxcs/Candle.
We propose to use automatically generated instruction-following data to improve the zero-shot capabilities of a large multimodal model with additional support for generative and image editing tasks. We achieve this by curating a new multimodal instruction-following set using GPT-4V and existing datasets for image generation and editing. Using this instruction set and the existing LLaVA-Finetune instruction set for visual understanding tasks, we produce GenLLaVA, a Generative Large Language and Visual Assistant. GenLLaVA is built through a strategy that combines three types of large pretrained models through instruction finetuning: Mistral for language modeling, SigLIP for image-text matching, and StableDiffusion for text-to-image generation. Our model demonstrates visual understanding capabilities superior to LLaVA and additionally demonstrates competitive results with native multimodal models such as Unified-IO 2, paving the way for building advanced general-purpose visual assistants by effectively re-using existing multimodal models. We open-source our dataset, codebase, and model checkpoints to foster further research and application in this domain.
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.
Due to high labeling cost, it is inevitable to introduce a certain proportion of noisy correspondence into visual-text datasets, resulting in poor model robustness for cross-modal matching. Although recent methods divide the datasets into clean and noisy pair subsets to yield promising achievements, they still suffer from deep neural networks over-fitting on noisy correspondence. In particular, the similar positive pairs with partially relevant semantic correspondence are easily partitioned into noisy pair subset by mistake without carefully selection, which brings harmful impact on robust learning. Meanwhile, the similar negative pairs with partially relevant semantic correspondence lead to ambiguous distance relations in common space learning, which also damages the stability of performance. To solve the coarse-grained dataset division problem, we propose Correspondence Tri-Partition Rectifier (CTPR) to partition the training set into clean, hard, and noisy pair subsets based on the memorization effect of neural networks and prediction inconsistency. Then, we refine the correspondence labels for each subset to indicate the real semantic correspondence between visual-text pairs. The differences between rectified labels of anchors and hard negatives are recast as the adaptive margin in the improved triplet loss for robust training in a co-teaching manner. To verify the effectiveness and robustness of our method, we conduct experiments by implementing image-text and video-text matching as two showcases. Extensive experiments on Flickr30 K, MS-COCO, MSR-VTT, and LSMDC datasets verify that our method successfully partitions the visual-text pairs according to their semantic correspondence and improves performance under noisy data training.
Image-Text Matching (ITM) aims to establish the correspondence between images and sentences. ITM is fundamental to various vision and language understanding tasks. However, there are limitations in the way existing ITM benchmarks are constructed. The ITM benchmark collects pairs of images and sentences during construction. Therefore, only samples that are paired at collection are annotated as positive. All other samples are annotated as negative. Many correlations are missed in these samples that are annotated as negative. For example, a sentence matches only one image at the time of collection. Only this image is annotated as positive for the sentence. All other images are annotated as negative. However, these negative images may contain images that correspond to the sentences. These mislabeled samples are called false negatives . Existing ITM models are optimized based on annotations containing mislabels, which can introduce noise during training. In this paper, we propose an ITM framework integrating Language Guidance ( LG ) for correcting false negatives. A language pre-training model is introduced into the ITM framework to identify false negatives. To correct false negatives, we propose language guidance loss, which adaptively corrects the locations of false negatives in the visual-semantic embedding space. Extensive experiments on two ITM benchmarks show that our method can improve the performance of existing ITM models. To verify the performance of correcting false negatives, we conduct further experiments on ECCV Caption. ECCV Caption is a verified dataset where false negatives in annotations have been corrected. The experimental results show that our method can recall more relevant false negatives.
Fact-checking is a crucial natural language processing (NLP) task that verifies the truthfulness of claims by considering reliable evidence. Traditional methods are often limited by labour-intensive data curation and rule-based approaches. In this paper, we present FactGenius, a novel method that enhances fact-checking by combining zero-shot prompting of large language models (LLMs) with fuzzy text matching on knowledge graphs (KGs). Leveraging DBpedia, a structured linked data dataset derived from Wikipedia, FactGenius refines LLM-generated connections using similarity measures to ensure accuracy. The evaluation of FactGenius on the FactKG, a benchmark dataset for fact verification, demonstrates that it significantly outperforms existing baselines, particularly when fine-tuning RoBERTa as a classifier. The two-stage approach of filtering and validating connections proves crucial, achieving superior performance across various reasoning types and establishing FactGenius as a promising tool for robust fact-checking. The code and materials are available at https://github.com/SushantGautam/FactGenius.
Semantic relevance calculation is crucial for e-commerce search engines, as it ensures that the items selected closely align with customer intent. Inadequate attention to this aspect can detrimentally affect user experience and engagement. Traditional text-matching techniques are prevalent but often fail to capture the nuances of search intent accurately, so neural networks now have become a preferred solution to processing such complex text matching. Existing methods predominantly employ representation-based architectures, which strike a balance between high traffic capacity and low latency. However, they exhibit significant shortcomings in generalization and robustness when compared to interaction-based architectures. In this work, we introduce a robust interaction-based modeling paradigm to address these shortcomings. It encompasses 1) a dynamic length representation scheme for expedited inference, 2) a professional terms recognition method to identify subjects and core attributes from complex sentence structures, and 3) a contrastive adversarial training protocol to bolster the model's robustness and matching capabilities. Extensive offline evaluations demonstrate the superior robustness and effectiveness of our approach, and online A/B testing confirms its ability to improve relevance in the same exposure position, resulting in more clicks and conversions. To the best of our knowledge, this method is the first interaction-based approach for large e-commerce search relevance calculation. Notably, we have deployed it for the entire search traffic on alibaba.com, the largest B2B e-commerce platform in the world.
Video moment retrieval (VMR) is to search for a visual temporal moment in an untrimmed raw video by a given text query description (sentence). Existing studies either start from collecting exhaustive frame-wise annotations on the temporal boundary of target moments (fully-supervised), or learn with only the video-level video-text pairing labels (weakly-supervised). The former is poor in generalisation to unknown concepts and/or novel scenes due to restricted dataset scale and diversity under expensive annotation costs; the latter is subject to visual-textual mis-correlations from incomplete labels. In this work, we introduce a new approach called hybrid-learning video moment retrieval to solve the problem by knowledge transfer through adapting the video-text matching relationships learned from a fully-supervised source domain to a weakly-labelled target domain when they do not share a common label space. Our aim is to explore shared universal knowledge between the two domains in order to improve model learning in the weakly-labelled target domain. Specifically, we introduce a multiplE branch Video-text Alignment model (EVA) that performs cross-modal (visual-textual) matching information sharing and multi-modal feature alignment to optimise domain-invariant visual and textual features as well as per-task discriminative joint video-text representations. Experiments show EVA's effectiveness in exploring temporal segment annotations in a source domain to help learn video moment retrieval without temporal labels in a target domain.
Memorization in large language models (LLMs) is a growing concern. LLMs have been shown to easily reproduce parts of their training data, including copyrighted work. This is an important problem to solve, as it may violate existing copyright laws as well as the European AI Act. In this work, we propose a systematic analysis to quantify the extent of potential copyright infringements in LLMs using European law as an example. Unlike previous work, we evaluate instruction-finetuned models in a realistic end-user scenario. Our analysis builds on a proposed threshold of 160 characters, which we borrow from the German Copyright Service Provider Act and a fuzzy text matching algorithm to identify potentially copyright-infringing textual reproductions. The specificity of countermeasures against copyright infringement is analyzed by comparing model behavior on copyrighted and public domain data. We investigate what behaviors models show instead of producing protected text (such as refusal or hallucination) and provide a first legal assessment of these behaviors. We find that there are huge differences in copyright compliance, specificity, and appropriate refusal among popular LLMs. Alpaca, GPT 4, GPT 3.5, and Luminous perform best in our comparison, with OpenGPT-X, Alpaca, and Luminous producing a particularly low absolute number of potential copyright violations. Code can be found at https://github.com/felixbmuller/llms-memorization-copyright.
Image-text matching has been a long-standing problem, which seeks to connect vision and language through semantic understanding. Due to the capability to manage large-scale raw data, unsupervised hashing-based approaches have gained prominence recently. They typically construct a semantic similarity structure using the natural distance, which subsequently provides guidance to the model optimization process. However, the similarity structure could be biased at the boundaries of semantic distributions, causing error accumulation during sequential optimization. To tackle this, we introduce a novel hashing approach termed Distribution-based Structure Mining with Consistency Learning (DEMO) for efficient image-text matching. From a statistical view, DEMO characterizes each image using multiple augmented views, which are considered as samples drawn from its intrinsic semantic distribution. Then, we employ a non-parametric distribution divergence to ensure a robust and precise similarity structure. In addition, we introduce collaborative consistency learning which not only preserves the similarity structure in the Hamming space but also encourages consistency between retrieval distribution from different directions in a self-supervised manner. Through extensive experiments on three benchmark image-text matching datasets, we demonstrate that DEMO achieves superior performance compared with many state-of-the-art methods.
The Learning-to-match (LTM) framework proves to be an effective inverse optimal transport approach for learning the underlying ground metric between two sources of data, facilitating subsequent matching. However, the conventional LTM framework faces scalability challenges, necessitating the use of the entire dataset each time the parameters of the ground metric are updated. In adapting LTM to the deep learning context, we introduce the mini-batch Learning-to-match (m-LTM) framework for audio-text retrieval problems. This framework leverages mini-batch subsampling and Mahalanobis-enhanced family of ground metrics. Moreover, to cope with misaligned training data in practice, we propose a variant using partial optimal transport to mitigate the harm of misaligned data pairs in training data. We conduct extensive experiments on audio-text matching problems using three datasets: AudioCaps, Clotho, and ESC-50. Results demonstrate that our proposed method is capable of learning rich and expressive joint embedding space, which achieves SOTA performance. Beyond this, the proposed m-LTM framework is able to close the modality gap across audio and text embedding, which surpasses both triplet and contrastive loss in the zero-shot sound event detection task on the ESC-50 dataset. Notably, our strategy of employing partial optimal transport with m-LTM demonstrates greater noise tolerance than contrastive loss, especially under varying noise ratios in training data on the AudioCaps dataset. Our code is available at https://github.com/v-manhlt3/m-LTM-Audio-Text-Retrieval