ACM MM 2024 Papers

Large language model (LLM) based knowledge graph completion (KGC) aims to predict the missing triples in the KGs with LLMs. However, research about LLM-based KGC fails to sufficiently harness LLMs' inference proficiencies, overlooking critical structural information integral to KGs. In this paper, we explore methods to incorporate structural information into the LLMs, with the overarching goal of facilitating structure-aware reasoning. We first discuss on the existing LLM paradigms like in-context learning and instruction tuning, proposing basic structural information injection approaches. Then we propose a Knowledge Prefix Adapter (KoPA) to fulfill this stated goal. KoPA uses a structural pre-training phase to comprehend the intricate entities and relations within KGs, representing them as structural embeddings. Then KoPA communicates such cross-modal structural information understanding to the LLMs through a knowledge prefix adapter which projects the structural embeddings into the textual space and obtains virtual knowledge tokens positioned as a prefix of the input prompt. We conduct comprehensive experiments and provide incisive analysis. Our code and data are available at https://github.com/zjukg/KoPA.

Video Temporal Grounding (VTG) aims to identify visual frames in a video clip that match text queries. Recent studies in VTG employ cross-attention to correlate visual frames and text queries as individual token sequences. However, these approaches overlook a crucial aspect of the problem: a holistic understanding of the query sentence. A model may capture correlations between individual word tokens and arbitrary visual frames while possibly missing out on the global meaning. To address this, we introduce two primary contributions: (1) a visual frame-level gate mechanism that incorporates holistic textual information, (2) cross-modal alignment loss to learn the fine-grained correlation between query and relevant frames. As a result, we regularize the effect of individual word tokens and suppress irrelevant visual frames. We demonstrate that our method outperforms state-of-the-art approaches in VTG benchmarks, indicating that holistic text understanding guides the model to focus on the semantically important parts within the video.

Large multimodal models (LMMs) have shown remarkable performance in the visual commonsense reasoning (VCR) task, which aims to answer a multiple-choice question based on visual commonsense within an image. However, the ability of LMMs to correct potential visual commonsense errors in the distractor upon their occurrence is yet under-explored. Drawing inspiration from how a human teacher crafts challenging distractors to test students' comprehension of the concepts or skills and assists them in identifying and correcting errors toward the answer, we are the pioneering research for LMMs to simulate this error correction process. To this end, we employ GPT-4 as a ''teacher'' to collect the explainable feedback dataset VCR-DF for error correction, which serves as a benchmark to evaluate the ability of LMMs to identify misconceptions and clarify reasons behind the error in VCR distractors toward final answers. In addition, we propose an LMM-based Pedagogical Expert Instructed Feedback Generation (PEIFG) model to incorporate the learnable expert prompts and multimodal instruction as guidance for feedback generation. Experimental results show that our PEIFG significantly outperforms existing LMMs. We believe that our benchmark provides a new direction for evaluating the capabilities of LMMs.

Audio Identification aims to precisely retrieve exact matches from a vast music repository through a query audio snippet. The need for specificity and granularity has traditionally led to representing music audio using numerous short fixed-duration overlapped segment/shingle features in fingerprinting approaches. However, fingerprinting imposes constraints on scalability and efficiency, as hundreds or even thousands of embeddings are generated to represent a music audio. In this paper, we present an innovative self-supervised approach called Angular Margin Guided Embedding (AMG-Embedding). AMG-Embedding is built on a traditional fingerprinting encoder and aims to represent variable-duration non-overlapped segments as embeddings through a two-stage embedding and class-level learning process. AMG-Embedding significantly reduces the number of generated embeddings while achieving high-specific fragment-level audio identification simultaneously. Experimental results demonstrate that AMG-Embedding achieves retrieval accuracy comparable to the based fingerprinting approach while consuming less than 1/10th of its storage and retrieval time. The efficiency gains of our approach position it as a promising solution for scalable and efficient audio identification systems.

The 3D model can be estimated by regressing the pose and shape parameters from the image data of the digital model. The reconstruction of 3D cartoon characters poses a challenging task due to diverse visual representations and postural variations. This paper proposes a dual-branch structure named MagicCartoon for 3D bipedal cartoon character estimation, which models pose and shape independently through feature decoupling. Considering the correlation between category difference and shape parameters, a hybrid feature fusion technique is introduced, which integrates the global features of the original image with the corresponding local features expressed by the puzzle image, reducing the abstractness of understanding shape parameter differences. To semantically align image and geometric between feature space, a geometric-guided feedback loop is proposed in an iterative way, so that the pose of modeling results can be expressed consistently with the image. Moreover, a feature consistency loss is designed to augment the training data by incorporating the same character with different postures and the same posture of different characters. It enhances the correlation between the features extracted by the backbone network and the specific task. Experiments conducted on the 3DBiCar dataset demonstrate that MagicCartoon outperforms the state-of-the-art methods.

In this work, borrowing a solution from the large-scale vision-language models (VLMs) instead of directly removing modality-specific signals from visual features, we propose a novel Flexible Modal CLIP (FM-CLIP) for flexible modal FAS, that can utilize text features to dynamically adjust visual features to be modality independent. In the visual branch, considering the huge visual differences of the same attack in different modalities, which makes it difficult for classifiers to flexibly identify subtle spoofing clues in different test modalities, we propose Cross-Modal Spoofing Enhancer (CMS-Enhancer). It includes a Frequency Extractor (FE) and Cross-Modal Interactor (CMI), aiming to map different modal attacks in a shared frequency space to reduce interference from modality-specific signals and enhance spoofing clues by leveraging cross-modal learning from the shared frequency space. In the text branch, we introduce a Language-Guided Patch Alignment (LGPA) based on prompt learning, which further guides the image encoder to focus on patch-level spoofing representations through dynamic weighting by text features. Thus, our FM-CLIP can flexibly test different modal samples by identifying and enhancing modality-agnostic spoofing cues. Finally, extensive experiments show that FM-CLIP is effective and outperforms state-of-the-art methods on multiple multi-modal datasets.

The increasing use of smartphones for capturing documents in various real-world conditions has underscored the need for robust document localization technologies. Current challenges in this domain include handling diverse document types, complex backgrounds, and varying photographic conditions such as low contrast and occlusion. However, there currently are no publicly available datasets containing these complex scenarios and few methods demonstrate their capabilities on these complex scenes. To address these issues, we create a new comprehensive real-world document localization benchmark dataset which contains the complex scenarios mentioned above and propose a novel Real-world Document Localization Network (RDLNet) for locating targeted documents in the wild. The RDLNet consists of an innovative light-SAM encoder and a masked attention decoder. Utilizing light-SAM encoder, the RDLNet transfers the mighty generalization capability of SAM to the document localization task. In the decoding stage, the RDLNet exploits the masked attention and object query method to efficiently output the triple-branch predictions consisting of corner point coordinates, instance-level segmentation area and categories of different documents without extra post-processing. We compare the performance of RDLNet with other state-of-the-art approaches for real-world document localization on multiple benchmarks, the results of which reveal that the RDLNet remarkably outperforms contemporary methods, demonstrating its superiority in terms of both accuracy and practicability.

Visual effects synthesis is crucial in the film and television industry, which aims at enhancing raw footage with virtual elements for greater expressiveness. As the demand for detailed and realistic effects escalates in modern production, professionals are compelled to allocate substantial time and resources to this endeavor. Thus, there is an urgent need to explore more convenient and less resource-intensive methods, such as incorporating the burgeoning Artificial Intelligence Generated Content (AIGC) technology. However, research into this potential integration has yet to be conducted. As the first work to establish a connection between visual effects synthesis and AIGC technology, we start by carefully setting up two paradigms according to the need for pre-produced effects or not: synthesis with reference effects and synthesis without reference effects. Following this, we compile a dataset by processing a collection of effects videos and scene videos, which contains a wide variety of effect categories and scenarios, adequately covering the common effects seen in films and television industry. Furthermore, we explore the capabilities of a pre-trained text-to-video model to synthesize visual effects within these two paradigms. The experimental results demonstrate that the pipeline we established can effectively produce impressive visual effects synthesis outcomes, thereby evidencing the significant potential of existing AIGC technology for application in visual effects synthesis tasks. Our dataset can be found in https://github.com/ruffiann/MagicVFX.

In the field of affective computing, fully leveraging information from a variety of sensory modalities is essential for the comprehensive understanding and processing of human emotions. Inspired by the process through which the human brain handles emotions and the theory of cross-modal plasticity, we propose UMBEnet, a brain-like unified modal affective processing network. The primary design of UMBEnet includes a Dual-Stream (DS) structure that fuses inherent prompts with a Prompt Pool and a Sparse Feature Fusion (SFF) module. The design of the Prompt Pool is aimed at integrating information from different modalities, while inherent prompts are intended to enhance the system's predictive guidance capabilities and effectively manage knowledge related to emotion classification. Moreover, considering the sparsity of effective information across different modalities, the SSF module aims to make full use of all available sensory data through the sparse integration of modality fusion prompts and inherent prompts, maintaining high adaptability and sensitivity to complex emotional states. Extensive experiments on the largest benchmark datasets in the Dynamic Facial Expression Recognition (DFER) field, including DFEW, FERV39k, and MAFW, have proven that UMBEnet consistently outperforms the current state-of-the-art methods. Notably, in scenarios of Modality Missingness and multimodal contexts, UMBEnet significantly surpasses the leading current methods, demonstrating outstanding performance and adaptability in tasks that involve complex emotional understanding with rich multimodal information. Code can be obtained at https://github.com/Xinji-Mai/UMBEnet.

Multimodal knowledge graph (MKG) reasoning has attracted significant attention since impressive performance has been achieved by adding multimodal auxiliary information (i.e., texts and images) to the entities of traditional KGs. However, existing studies heavily rely on path-based methods for learning structural modality, failing to capture the complex structural interactions among multimodal entities beyond the reasoning path. In addition, existing studies have largely ignored the dynamic impact of different multimodal features on different decision facts for reasoning, which utilize asymmetric coattention to independently learn the static interplay between different modalities without dynamically joining the reasoning process. We propose a novel Dynamic Structure-aware representation learning method, namely DySarl, to overcome this problem and significantly improve the MKG reasoning performance. Specifically, we devise a dual-space multihop structural learning module in DySarl, aggregating the multihop structural features of multimodal entities via a novel message-passing mechanism. It integrates the message paradigms in Euclidean and hyperbolic spaces, effectively preserving the neighborhood information beyond the limited multimodal query paths. Furthermore, DySarl has an interactive symmetric attention module to explicitly learn the dynamic impacts of unimodal attention senders and multimodal attention targets on decision facts through a newly designed symmetric attention component and fact-specific gated attention unit, equipping DySarl with the dynamic associations between the multimodal feature learning and later reasoning. Extensive experiments show that DySarl achieves significantly improved reasoning performance on two public MKG datasets compared with that of the state-of-the-art baselines. Source codes are available at https://github.com/HUSTNLP-codes/DySarl.

Unsupervised visible-infrared person re-identification (USL-VI-ReID) is of great research and practical significance yet remains challenging due to significant modality discrepancy and lack of annotations. Many existing approaches utilize variants of bipartite graph global matching algorithms to address this issue, aiming to establish cross-modality correspondences. However, these methods may encounter mismatches due to significant modality gaps and limited model representation. To mitigate this, we propose a simple yet effective framework for USL-VI-ReID, which gradually establishes associations between different modalities. To measure the confidence whether samples from different modalities belong to the same identity, we introduce a bidirectional-consistency criterion, which not only considers direct relationships between samples from different modalities but also incorporates potential hard negative samples from the same modality. Additionally, we propose a cross-modality correlation preserving module to further enhance the semantic representation of the model by maintaining consistency in correlations across modalities. Extensive experiments conducted on the public SYSU-MM01 and RegDB datasets demonstrate the superiority of our method over existing USL-VI-ReID approaches across various settings, despite the simplicity of our method.

Advances in multimedia technology and its widespread application in education have made multimedia learning increasingly important. Knowledge Tracing (KT) is the key technology for achieving adaptive multimedia learning, aiming to monitor the degree of knowledge acquisition and predict students' performance during the learning process. Current KT research is dedicated to enhancing the performance of KT problems by integrating the most advanced deep learning techniques. However, this has led to increasingly complex models, which reduce model usability and divert researchers' attention away from exploring the core issues of KT. This paper aims to tackle the fundamental challenges of KT tasks, including the knowledge state representation and the core architecture design, and investigate a novel KT model that is both simple and powerful. We have revisited the KT task and propose the ReKT model. First, taking inspiration from the decision-making process of human teachers, we model the knowledge state of students from three distinct perspectives: questions, concepts, and domains. Second, building upon human cognitive development models, such as constructivism, we have designed a Forget-Response-Update (FRU) framework to serve as the core architecture for the KT task. The FRU is composed of just two linear regression units, making it an extremely lightweight framework. Extensive comparisons were conducted with 22 state-of-the-art KT models on 7 publicly available datasets. The experimental results demonstrate that ReKT outperforms all the comparative methods in question-based KT tasks, and consistently achieves the best (in most cases) or near-best performance in concept-based KT tasks. Furthermore, in comparison to other KT core architectures like Transformers or LSTMs, the FRU achieves superior prediction performance with approximately only 38% computing resources. Through an exploration of the ReKT model that is both simple and powerful, is able to offer new insights to future KT research. The code can be found at https://github.com/lilstrawberry/ReKT.

With the continuous development of imaging technology and the gradual expansion of the amount of image data, how to achieve high compression efficiency of high-resolution images is a challenge problem for storage and transmission. Image rescaling aims to reduce the original data amount through downscaling to facilitate data transmission and storage before encoding, and reconstruct the quality through upscaling after decoding, which is a key technology to assist in high-ratio image compression. However, existing rescaling approaches are more focused on reconstruction quality rather than image compressibility. In repetitive observation scenarios, multi-temporal images brought by periodic observations provide an opportunity to alleviate the conflict between reconstruction quality and compressibility, that is, the historical images as reference indicates what information can be dropped at downscaling to reduce the information content in downscaled image and provides the dropped information to improve the image restoration quality at upscaling. Based on this consideration, we propose a novel multi-temporal assisted reference-based image rescaling framework (RefScale). Specifically, a referencing network is proposed to calculate the similarity map to provide the referencing condition, which is then injected into the conditional invertible neural network to guide the information drop at the downscaling stage and information fusion at the upscaling stage. Additionally, a low-resolution guidance loss is proposed to further constrain the data amount of the downscaled image. Experiments conducted on both satellite imaging and autonomous driving show the superior performance of our approach over the state-of-the-art methods.

Diffusion-based text-to-image personalization have achieved great success in generating subjects specified by users among various contexts. Even though, existing finetuning-based methods still suffer from model overfitting, which greatly harms the generative diversity, especially when given subject images are few. To this end, we propose Pick-and-Draw, a training-free semantic guidance approach to boost identity consistency and generative diversity for personalization methods. Our approach consists of two components: appearance picking guidance and layout drawing guidance. As for the former, we construct an appearance palette with visual features from the reference image, where we pick local patterns for generating the specified subject with consistent identity. As for layout drawing, we outline the subject's contour by referring to a generative template from the vanilla diffusion model, and inherit the strong image prior to synthesize diverse contexts according to different text conditions. The proposed approach can be applied to any personalized diffusion models and requires as few as a single reference image. Qualitative and quantitative experiments show that Pick-and-Draw consistently improves identity consistency and generative diversity, pushing the trade-off between subject fidelity and image-text fidelity to a new Pareto frontier.

In challenging low-light and adverse weather conditions, thermal vision algorithms, especially object detection, have exhibited remarkable potential, contrasting with the frequent struggles encountered by visible vision algorithms. Nevertheless, the efficacy of thermal vision algorithms driven by deep learning models remains constrained by the paucity of available training data samples. To this end, this paper introduces a novel approach termed the edge-guided conditional diffusion model (ECDM). This framework aims to produce meticulously aligned pseudo thermal images at the pixel level, leveraging edge information extracted from visible images. By utilizing edges as contextual cues from the visible domain, the diffusion model achieves meticulous control over the delineation of objects within the generated images. To alleviate the impacts of those visible-specific edge information that should not appear in the thermal domain, a two-stage modality adversarial training (TMAT) strategy is proposed to filter them out from the generated images by differentiating the visible and thermal modality. Extensive experiments on LLVIP demonstrate ECDM's superiority over existing state-of-the-art approaches in terms of image generation quality. The pseudo thermal images generated by ECDM also help to boost the performance of various thermal object detectors by up to 7.1 mAP. Code is available at https://github.com/lengmo1996/ECDM.

With the rapid advancements of large-scale text-to-image diffusion models, various practical applications have emerged, bringing significant convenience to society. However, model developers may misuse the unauthorized data to train diffusion models. These data are at risk of being memorized by the models, thus potentially violating citizens' privacy rights. Therefore, in order to judge whether a specific image is utilized as a member of a model's training set, Membership Inference Attack (MIA) is proposed to serve as a tool for privacy protection. Current MIA methods predominantly utilize pixel-wise comparisons as distinguishing clues, considering the pixel-level memorization characteristic of diffusion models. However, it is practically impossible for text-to-image models to memorize all the pixel-level information in massive training sets. Therefore, we move to the more advanced structure-level memorization. Observations on the diffusion process show that the structures of members are better preserved compared to those of nonmembers, indicating that diffusion models possess the capability to remember the structures of member images from training sets. Drawing on these insights, we propose a simple yet effective MIA method tailored for text-to-image diffusion models. Extensive experimental results validate the efficacy of our approach. Compared to current pixel-level baselines, our approach not only achieves state-of-the-art performance but also demonstrates remarkable robustness against various distortions.

Industrial multimedia recommendation systems extensively utilize cascade architectures to deliver personalized content for users, generally consisting of multiple stages like retrieval and ranking. However, retrieval models have long suffered from Sample Selection Bias (SSB) due to the distribution discrepancy between the exposed items used for model training and the candidates (almost unexposed) during inference, affecting recommendation performance. Traditional methods utilize retrieval candidates as augmented training data, indiscriminately treating unexposed data as negative samples, which leads to inaccuracies and noise. Some efforts rely on unbiased datasets, while they are costly to collect and insufficient for industrial models. In this paper, we propose a debiasing framework named DAMCAR, which introduces Domain Adaptation to mitigate SSB in Multimedia CAscade Recommendation systems. Firstly, we sample hard-to-distinguish samples from unexposed data to serve as the target domain, optimizing data quality and resource utilization. Secondly, adversarial domain adaptation is employed to generate pseudo-labels for each sample. To enhance robustness, we utilize Exponential Moving Average (EMA) to create a teacher model that supervises the generation of pseudo-labels via self-distillation. Finally, we obtain a retrieval model that maintains stable performance during inference through a hybrid training mechanism. We conduct offline experiments on two real-world datasets and deploy our approach in the retrieval model of a multimedia video recommendation system for online A/B testing. Comprehensive experimental results demonstrate the effectiveness of DAMCAR in practical applications.