There are various factors affecting the performance of video search. An imprecise query will enlarge search space and reduce the discriminative power of ranking functions. This problem is further exacerbated by the presence of numerous visually or semantically similar videos in large datasets. Consequently, users need to painstakingly browse through many highly similar candidates to locate the search target, leading to increased cognitive load and inefficient searching. Ideally, engaging users through interactive questioning to resolve uncertainties in the search process is an effective strategy for progressively narrowing down the search space. However, despite rapid advances in deep learning, generating informative questions conditioned on the user query and search result remains a highly difficult problem.

Large Multi-modal Models (LMMs) have significantly advanced a variety of vision-language tasks. The scalability and availability of high-quality training data play a pivotal role in the success of LMMs. In the realm of food, while comprehensive food datasets such as Recipe1M offer an abundance of ingredient and recipe information, they often fall short of providing ample data for nutritional analysis. The Recipe1M+ dataset, despite offering a subset for nutritional evaluation, is limited in the scale and accuracy of nutrition information. To bridge this gap, we introduce Uni-Food, a unified food dataset that comprises over 100,000 images with various food labels, including categories, ingredients, recipes, and ingredient-level nutritional information. To mitigate the conflicts arising from multi-task supervision during fine-tuning of LMMs, we introduce a novel Linear Rectification Mixture of Diverse Experts (RoDE) approach. RoDE utilizes a diverse array of experts to address tasks of varying complexity, thereby facilitating the coordination of trainable parameters, i.e., it allocates more parameters for more complex tasks and, conversely, fewer parameters for simpler tasks. RoDE implements linear rectification union to refine the router’s functionality, thereby enhancing the efficiency of sparse task allocation. These design choices endow RoDE with features that ensure GPU memory efficiency and ease of optimization. Extensive experiments validate the effectiveness of our approach in addressing the inherent challenges of food-related multitasking. UniFood Project

Multimodal Large Language Models (MLLMs) have exhibited remarkable advancements in integrating different modalities, excelling in complex understanding and generation tasks. Despite their success, MLLMs remain vulnerable to conversational adversarial inputs. In this paper, we systematically study gaslighting negation attacks—a phenomenon where models, despite initially providing correct answers, are persuaded by user-provided negations to reverse their outputs, often fabricating justifications. We conduct extensive evaluations of state-of-the-art MLLMs across diverse benchmarks and observe substantial performance drops when negation is introduced. Notably, we introduce the first benchmark GaslightingBench, specifically designed to evaluate the vulnerability of MLLMs to negation arguments. GaslightingBench consists of multiple-choice questions curated from existing datasets, along with generated negation prompts across 20 diverse categories. Throughout extensive evaluation, we find that proprietary models such as Gemini-1.5-flash and GPT-4o demonstrate better resilience compared to open-source counterparts like Qwen2-VL and LLaVA, though even advanced reasoning-oriented models like Gemini-2.5-Pro remain susceptible. Our category-level analysis further shows that subjective or socially nuanced domains (e.g., Social Relation, Image Emotion) are especially fragile, while more objective domains (e.g., Geography) exhibit relatively smaller but still notable drops. Overall, all evaluated MLLMs struggle to maintain logical consistency under gaslighting negation attack. These findings highlight a fundamental robustness gap and provide insights for developing more reliable and trustworthy multimodal AI systems. Project website: https://yxg1005.github.io/GaslightingNegationAttacks.

Known-item search (KIS) involves only a single search target, making relevance feedback-typically a powerful technique for efficiently identifying multiple positive examples to infer user intent-inapplicable. PicHunter addresses this issue by asking users to select the top-k most similar examples to the unique search target from a displayed set. Under ideal conditions, when the user's perception aligns closely with the machine's perception of similarity, consistent and precise judgments can elevate the target to the top position within a few iterations. However, in practical scenarios, expecting users to provide consistent judgments is often unrealistic, especially when the underlying embedding features used for similarity measurements lack interpretability. To enhance robustness, we first introduce a pairwise relative judgment feedback that improves the stability of top-k selections by mitigating the impact of misaligned feedback. Then, we decompose user perception into multiple sub-perceptions, each represented as an independent embedding space. This approach assumes that users may not consistently align with a single representation but are more likely to align with one or several among multiple representations. We develop a predictive user model that estimates the combination of sub-perceptions based on each user feedback instance. The predictive user model is then trained to filter out the misaligned sub-perceptions. Experimental evaluations on the large-scale open-domain dataset V3C indicate that the proposed model can optimize over 60% search targets to the top rank when their initial ranks at the search depth between 10 and 50. Even for targets initially ranked between 1,000 and 5,000, the model achieves a success rate exceeding 40% in optimizing ranks to the top, demonstrating the enhanced robustness of relevance feedback in KIS despite inconsistent feedback.

Nutrition estimation is an important component of promoting healthy eating and mitigating diet-related health risks. Despite advances in tasks such as food classification and ingredient recognition, progress in nutrition estimation is limited due to the lack of datasets with nutritional annotations. To address this issue, we introduce FastFood, a dataset with 84,446 images across 908 fast food categories, featuring ingredient and nutritional annotations. In addition, we propose a new model-agnostic Visual-Ingredient Feature Fusion (VIF2) method to enhance nutrition estimation by integrating visual and ingredient features. Ingredient robustness is improved through synonym replacement and resampling strategies during training. The ingredient-aware visual feature fusion module combines ingredient features and visual representation to achieve accurate nutritional prediction. During testing, ingredient predictions are refined using large multimodal models by data augmentation and majority voting. Our experiments on both FastFood and Nutrition5k datasets validate the effectiveness of our proposed method built in different backbones (e.g., Resnet, InceptionV3 and ViT), which demonstrates the importance of ingredient information in nutrition estimation.

Aligning a user query and video clips in cross-modal latent space and that with semantic concepts are two mainstream approaches for ad-hoc video search (AVS). However, the effectiveness of existing approaches is bottlenecked by the small sizes of available video-text datasets and the low quality of concept banks, which results in the failures of unseen queries and the out-of-vocabulary problem. This paper addresses these two problems by constructing a new dataset and developing a multi-word concept bank. Specifically, capitalizing on a generative model, we construct a new dataset consisting of 7 million generated text and video pairs for pre-training. To tackle the out-of-vocabulary problem, we develop a multi-word concept bank based on syntax analysis to enhance the capability of a state-of-the- art interpretable AVS method in modelling relationships between query words. We also study the impact of current advanced features on the method. Experimental results show that the integration of the above-proposed elements doubles the R@1 performance of the AVS method on the MSRVTT dataset and improves the xinfAP on the TRECVid AVS query sets for 2016-2023 (eight years) by a margin from 2% to 77%, with an average about 20%. The code and model are available at https://github.com/nikkiwoo-gh/Improved-ITV.

Cross-modal representation learning has become a new normal for bridging the semantic gap between text and visual data. Learning modality agnostic representations in a continuous latent space, however, is often treated as a black-box data-driven training process. It is well known that the effectiveness of representation learning depends heavily on the quality and scale of training data. For video representation learning, having a complete set of labels that annotate the full spectrum of video content for training is highly difficult, if not impossible. These issues, black-box training and dataset bias, make representation learning practically challenging to be deployed for video understanding due to unexplainable and unpredictable results. In this article, we propose two novel training objectives, likelihood and unlikelihood functions, to unroll the semantics behind embeddings while addressing the label sparsity problem in training. The likelihood training aims to interpret semantics of embeddings beyond training labels, while the unlikelihood training leverages prior knowledge for regularization to ensure semantically coherent interpretation. With both training objectives, a new encoder-decoder network, which learns interpretable cross-modal representation, is proposed for ad-hoc video search. Extensive experiments on TRECVid and MSR-VTT datasets show that the proposed network outperforms several state-of-the-art retrieval models with a statistically significant performance margin.

Cross-modal recipe retrieval has attracted research attention in recent years, thanks to the availability of large-scale paired data for training. Nevertheless, obtaining adequate recipe-image pairs covering the majority of cuisines for supervised learning is difficult if not impossible. By transferring knowledge learnt from a data-rich cuisine to a data-scarce cuisine, domain adaptation sheds light on this practical problem. Nevertheless, existing works assume recipes in source and target domains are mostly originated from the same cuisine and written in the same language. This paper studies unsupervised domain adaptation for image-to-recipe retrieval, where recipes in source and target domains are in different languages. Moreover, only recipes are available for training in the target domain. A novel recipe mixup method is proposed to learn transferable embedding features between the two domains. Specifically, recipe mixup produces mixed recipes to form an intermediate domain by discretely exchanging the section(s) between source and target recipes. To bridge the domain gap, recipe mixup loss is proposed to enforce the intermediate domain to locate in the shortest geodesic path between source and target domains in the recipe embedding space. By using Recipe 1M dataset as source domain (English) and Vireo-FoodTransfer dataset as target domain (Chinese), empirical experiments verify the effectiveness of recipe mixup for cross-lingual adaptation in the context of image-to-recipe retrieval.

Restaurant dishes represent a significant portion of food that people consume in their daily life. While people are becoming health-conscious in their food intake, convenient restaurant food tracking becomes an essential task in wellness and fitness applications. Given the huge number of dishes (food categories) involved, it becomes extremely challenging for traditional food photo classification to be feasible in both algorithm design and training data availability. In this work, we present a demo that runs on restaurant dish images in a city of millions of residents and tens of thousand restaurants. We propose a rank-loss based convolutional neural network to optimize the image features representation. Context information such as GPS location of the recognition request is also used to further improve the performance. Our experimental results are highly promising. We have shown in our demo that the proposed algorithm is near ready to be deployed in real-world applications.

There has been a tremendous growth in video data the last decade. People are using mobile phones and tablets to take, share or watch videos more than ever before. Video cameras are around us almost everywhere in the public domain (e.g. stores, streets, public facilities, ...etc). Efficient and effective retrieval methods are critically needed in different applications. The goal of TRECVID is to encourage research in content-based video retrieval by providing large test collections, uniform scoring procedures, and a forum for organizations interested in comparing their results. In this tutorial, we present and discuss some of the most important and fundamental content-based video retrieval problems such as recognizing predefined visual concepts, searching in videos for complex ad-hoc user queries, searching by image/video examples in a video dataset to retrieve specific objects, persons, or locations, detecting events, and finally bridging the gap between vision and language by looking into how can systems automatically describe videos in a natural language. A review of the state of the art, current challenges, and future directions along with pointers to useful resources will be presented by different regular TRECVID participating teams. Each team will present one of the following tasks:

A problem not well understood in video hyperlinking is what qualifies a fragment as an anchor or target. Ideally, anchors provide good starting points for navigation, and targets supplement anchors with additional details while not distracting users with irrelevant, false and redundant information. The problem is not trivial for intertwining relationship between data characteristics and user expectation. Imagine that in a large dataset, there are clusters of fragments spreading over the feature space. The nature of each cluster can be described by its size (implying popularity) and structure (implying complexity). A principle way of hyperlinking can be carried out by picking centers of clusters as anchors and from there reach out to targets within or outside of clusters with consideration of neighborhood complexity. The question is which fragments should be selected either as anchors or targets, in one way to reflect the rich content of a dataset, and meanwhile to minimize the risk of frustrating user experience. This paper provides some insights to this question from the perspective of hubness and local intrinsic dimensionality, which are two statistical properties in assessing the popularity and complexity of data space. Based these properties, two novel algorithms are proposed for low-risk automatic selection of anchors and targets.

Complex video event detection without visual examples is a very challenging issue in multimedia retrieval. We present a state-of-the-art framework for event search without any need of exemplar videos and textual metadata in search corpus. To perform event search given only query words, the core of our framework is a large, pre-built bank of concept detectors which can understand the content of a video in the perspective of object, scene, action and activity concepts. Leveraging such knowledge can effectively narrow the semantic gap between textual query and the visual content of videos. Besides the large concept bank, this paper focuses on two challenges that largely affect the retrieval performance when the size of the concept bank increases: (1) How to choose the right concepts in the concept bank to accurately represent the query; (2) if noisy concepts are inevitably chosen, how to minimize their influence. We share our novel insights on these particular problems, which paves the way for a practical system that achieves the best performance in NIST TRECVID 2015.

This demo showcases the utility of video hyperlinks with celebrities as the link anchors and their social circles as targets, aiming to help users quickly explore the aboutness of a celebrity by link traversal. Through content analysis, our system embeds hyperlinks into videos such that users can click-and-jump between celebrity faces in different videos to get-to-know their social circles. One peculiar feature is the ability of the system in providing links that maximize users' chance encounter, or serendipitous experience, beyond information need. Our system is enabled by two key components, name-face association and diversity-based ranking, for the aboutness and serendipity features respectively in hyperlinking. The former component assigns names video faces while mining celebrity social networks. The latter performs topic modeling so as to rank videos based on topic diversity.

This paper investigates the task of automatically associating faces appearing in images (or videos) with their names. Our novelty lies in the use of celebrity Web images to facilitate the task. Specifically, we first propose a method named Image Matching (IM), which uses the faces in images returned from name queries over an image search engine as the gallery set of the names, and a probe face is classified as one of the names, or none of them, according to their matching scores and compatibility characterized by a proposed Assigning-Thresholding (AT) pipeline. Noting IM could provide guidance for association for the well-established Graph-based Association (GA), we further propose two methods that jointly utilize the two kinds of complementary cues. They are: the early fusion of IM and GA (EF-IMGA) that takes the IM score as an additional information source to help the association in GA, and the late fusion of IM and GA (LF-IMGA) that combines the scores from both IM and GA obtained individually to make the association. Evaluations on datasets of captioned news images and Web videos both show the proposed methods, especially the two fused ones, provide significant improvements over GA.

Detecting emotions from user-generated videos, such as "anger" and "sadness", has attracted widespread interest recently. The problem is challenging as effectively representing video data with multi-view information (e.g., audio, video or text) is not trivial. In contrast to the existing works that extract features from each modality (view) separately followed by early or late fusion, we propose to learn a joint density model over the space of multi-modal inputs (including visual, auditory and textual modalities) with Deep Boltzmann Machine (DBM). The model is trained directly on the user-generated Web videos without any labeling effort. More importantly, the deep architecture enlightens the possibility of discovering the highly non-linear relationships that exist between low-level features across different modalities. The experiment results show that the DBM model learns joint representation complementary to the hand-crafted visual and auditory features, leading to 7.7% performance improvement in classification accuracy on the recently released VideoEmotion dataset.

Similarity search is one of the fundamental problems for large scale multimedia applications. Hashing techniques, as one popular strategy, have been intensively investigated owing to the speed and memory efficiency. Recent research has shown that leveraging supervised information can lead to high quality hashing. However, most existing supervised methods learn hashing function by treating each training example equally while ignoring the different semantic degree related to the label, i.e. semantic confidence, of different examples. In this paper, we propose a novel semi-supervised hashing framework by leveraging semantic confidence. Specifically, a confidence factor is first assigned to each example by neighbor voting and click count in the scenarios with label and click-through data, respectively. Then, the factor is incorporated into the pairwise and triplet relationship learning for hashing. Furthermore, the two learnt relationships are seamlessly encoded into semi-supervised hashing methods with pairwise and listwise supervision respectively, which are formulated as minimizing empirical error on the labeled data while maximizing the variance of hash bits or minimizing quantization loss over both the labeled and unlabeled data. In addition, the kernelized variant of semi-supervised hashing is also presented. We have conducted experiments on both CIFAR-10 (with label) and Clickture (with click data) image benchmarks (up to one million image examples), demonstrating that our approaches outperform the state-of-the-art hashing techniques.

In this demonstration, we demonstrate a mobile-based celebrity video browsing system called CeleBrowser. Using this system, users can interactively switch among four views: people-centric, timeline-centric, month-centric and topic-centric, for browsing celebrity-related hot videos. A peculiarity of the demonstration is to highlight the advantage of multi-perspective information organization and presentation in engaging users for exploratory browsing of large number of Web videos on a device with small screen. Technology-wise the demonstration shows how query logs collected for six months from two vertical search engines are leveraged for mining hot events and videos of celebrities.

One of the fundamental problems in image search is to rank image documents according to a given textual query. Existing search engines highly depend on surrounding texts for ranking images, or leverage the query-image pairs annotated by human labelers to train a series of ranking functions. However, there are two major limitations: 1) the surrounding texts are often noisy or too few to accurately describe the image content, and 2) the human annotations are resourcefully expensive and thus cannot be scaled up. We demonstrate in this paper that the above two fundamental challenges can be mitigated by jointly exploring the cross-view learning and the use of click-through data. The former aims to create a latent subspace with the ability in comparing information from the original incomparable views (i.e., textual and visual views), while the latter explores the largely available and freely accessible click-through data (i.e., ``crowdsourced" human intelligence) for understanding query. Specifically, we propose a novel cross-view learning method for image search, named Click-through-based Cross-view Learning (CCL), by jointly minimizing the distance between the mappings of query and image in the latent subspace and preserving the inherent structure in each original space. On a large-scale click-based image dataset, CCL achieves the improvement over Support Vector Machine-based method by 4.0\% in terms of relevance, while reducing the feature dimension by several orders of magnitude (e.g., from thousands to tens). Moreover, the experiments also demonstrate the superior performance of CCL to several state-of-the-art subspace learning techniques.

Instance Search (INS) is a realistic problem initiated by TRECVID, which is to retrieve all occurrences of the querying object, location, or person from a large video collection. It is a fundamental problem with many applications, and also a challenging problem different from the traditional concept or near-duplicate (ND) search, since the relevancy is defined at instance level. True responses could exhibit various visual variations, such as being small on the image with different background, or showing a non-homography spatial configuration. Based on the Bag-of-Words model, we propose two techniques tailored for Instance Search. Specifically, we explore the use of (1) an elastic spatial topology checking technique based on Delaunay Triangulation (DT), and (2) a practical background context modeling method by simulating the "stare" behavior of human eyes. With DT, we improve the quality of visual matching by accumulating evidence from local topology-preserving patches, significantly boosting the ranks of topology consistent results. On the other hand, we increase the information quantity for visual matching with the "stare" model, such that instances appearing in both similar and different background can be highly ranked as results. The proposed techniques are evaluated on the INS datasets of TRECVID, achieving large performance gain with small computation overhead, compared with several existing methods.

The search for entities is the most common search behavior on the Web, especially in social media communities where entities (such as images, videos, people, locations, and tags) are highly heterogeneous and correlated. While previous research usually deals with these social media entities separately, we are investigating in this paper a unified, multi-level, and correlative entity graph to represent the unstructured social media data, through which various applications (e.g., friend suggestion, personalized image search, image tagging, etc.) can be realized more effectively in one single framework. We regard the social media objects equally as "entities" and all of these applications as "entity search" problem which searches for entities with different types. We first construct a multi-level graph which organizes the heterogeneous entities into multiple levels, with one type of entities as vertices in each level. The edges between graphs pairwisely connect the entities weighted by intra-relations in the same level and inter-links across two different levels distilled from the social behaviors (e.g., tagging, commenting, and joining communities). To infer the strength of intra-relations, we propose a circular propagation scheme, which reinforces the mutual exchange of information across different entity types in a cyclic manner. Based on the constructed unified graph, we explicitly formulate entity search as a global optimization problem in a unified Bayesian framework, in which various applications are efficiently realized. Empirically, we validate the effectiveness of our unified entity graph for various social media applications on million-scale real-world dataset.

Multimedia documents in popular image and video sharing websites such as Flickr and Youtube are heterogeneous documents with diverse ways of representations and rich user-supplied information. In this paper, we investigate how the agreement among heterogeneous modalities can be exploited to guide data fusion. The problem of fusion is cast as the simultaneous mining of agreement from different modalities and adaptation of fusion weights to construct a fused graph from these modalities. An iterative framework based on agreement-fusion optimization is thus proposed. We plug in two well-known algorithms: random walk and semi-supervised learning to this framework to illustrate the idea of how agreement (conflict) is incorporated (compromised) in the case of uniform and adaptive fusion. Experimental results on web video and image re-ranking demonstrate that, by proper fusion strategy rather than simple linear fusion, performance improvement on search can generally be expected.

This paper presents an innovative style-wise advertising platform for web page. Web page mainly refers to visual effects, such as color and layout. Unlike the most popular ad-network such as Google AdSense which needs publishers to change the original structure of their pages and define the position and style of the embedded ads manually, style-wise page advertising aims to automatically deliver style-consistent ads at proper positions within the web page, without breaking the layout of the original page. Our system is motivated from the fact that almost 90% web pages contain blank regions without any content. Given a web page with some blank regions, style-wise page advertising is able to detect the suitable blank areas for advertising, rank the ads according to the semantic relevance and web page style, and embed relevant ads into these nonintrusive areas. Style-wise page advertising represents one of the first attempts towards contextual advertising which enables the publishers to save effort when applying online advertising services.

Bag-of-visual-words (BoW) has been popular for visual classification in recent years. In this paper, we propose a novel BoW expansion method to alleviate the effect of visual word correlation problem. We achieve this by diffusing the weights of visual words in BoW based on visual word relatedness, which is rigorously defined within a visual ontology. The proposed method is tested in video indexing experiment on TRECVID-2006 video retrieval benchmark, and an improvement of 7% over the traditional BoW is reported.