ACM MM 2024 Papers

Learning semantic-rich representations from unlabeled time series data with intricate dynamics is a notable challenge. Traditional contrastive learning techniques predominantly focus on segment-level augmentations through time slicing, a practice that, while valuable, often results in sampling bias and suboptimal performance due to the loss of global context. Furthermore, they typically disregard the vital frequency information to enrich data representations. To this end, we propose a novel self-supervised general-purpose framework called Temporal-Frequency and Contextual Consistency (TFCC). Specifically, this framework first performs two instance-level augmentation families over the entire series to capture nuanced representations alongside critical long-term dependencies. Then, TFCC advances by initiating dual cross-view forecasting tasks between the original series and its augmented counterpart in both time and frequency domains to learn robust representations. Finally, three specially designed consistency modules 'temporal, frequency, and temporal-frequency' aid in further developing discriminative representations on top of the learned robust representations. Extensive experiments on multiple benchmarks demonstrate TFCC's superiority over the state-of-the-art classification and forecasting methods and exhibit exceptional efficiency in semi-supervised and transfer learning scenarios.

Engaging in conversational recommendations within a specific scenario represents a promising paradigm in the real world. Scenario-relevant situations often affect conversations and recommendations from two closely related aspects: varying the appealingness of items to users, namely situated item representation, and shifting user interests in the targeted items, namely situated user preference. We highlight that considering those situational factors is crucial, as this aligns with the realistic conversational recommendation process in the physical world. However, it is challenging yet under-explored. In this work, we are pioneering to bridge this gap and introduce a novel setting: Situated Conversational Recommendation Systems (SCRS). We observe an emergent need for high-quality datasets, and building one from scratch requires tremendous human effort. To this end, we construct a new benchmark, named SCREEN, via a role-playing method based on multimodal large language models. We take two multimodal large language models to play the roles of a user and a recommender, simulating their interactions in a co-observed scene. Our SCREEN comprises over 20k dialogues across 1.5k diverse situations, providing a rich foundation for exploring situational influences on conversational recommendations. Based on the SCREEN, we propose three worth-exploring subtasks and evaluate several representative baseline models. Our evaluations suggest that the benchmark is high quality, establishing a solid experimental basis for future research. The code and data are available at https://github.com/DongdingLin/SCREEN.

Handling varying computational resources is a critical issue in modern AI applications. Adaptive deep networks, featuring the dynamic employment of multiple classifier heads among different layers, have been proposed to address classification tasks under varying computing resources. Existing approaches typically utilize the last classifier supported by the available resources for inference, as they believe that the last classifier always performs better across all classes. However, our findings indicate that earlier classifier heads can outperform the last head for certain classes. Based on this observation, we introduce the Collaborative Decision Making (CDM) module, which fuses the multiple classifier heads to enhance the inference performance of adaptive deep networks. CDM incorporates an uncertainty-aware fusion method based on evidential deep learning (EDL), that utilizes the reliability (uncertainty values) from the first c-1 classifiers to improve the c-th classifier' accuracy. We also design a balance term that reduces fusion saturation and unfairness issues caused by EDL constraints to improve the fusion quality of CDM. Finally, a regularized training strategy that uses the last classifier to guide the learning process of early classifiers is proposed to further enhance the CDM module's effect, called the Guided Collaborative Decision Making (GCDM) framework. The experimental evaluation demonstrates the effectiveness of our approaches. Results on ImageNet datasets show CDM and GCDM obtain 0.4% to 2.8% accuracy improvement (under varying computing resources) on popular adaptive networks. The code is available at the link https://github.com/Meteor-Stars/GCDM_AdaptiveNet.

Spiking neural networks (SNNs) show great potential due to their energy efficiency, fast processing capabilities, and robustness. There are two main approaches to constructing SNNs. Direct training methods require much memory, while conversion methods offer a simpler and more efficient option. However, current conversion methods mainly focus on converting convolutional neural networks (CNNs) to SNNs. Converting Transformers to SNN is challenging because of the presence of non-linear modules. In this paper, we propose an Expectation Compensation Module to preserve the accuracy of the conversion. The core idea is to use information from the previous T time-steps to calculate the expected output at time-step T. We also propose a Multi-Threshold Neuron and the corresponding Parallel Parameter normalization to address the challenge of large time steps needed for high accuracy, aiming to reduce network latency and power consumption. Our experimental results demonstrate that our approach achieves state-of-the-art performance. For example, we achieve a top-1 accuracy of 88.60% with only a 1% loss in accuracy using 4 time steps while consuming only 35% of the original power of the Transformer. To our knowledge, this is the first successful Artificial Neural Network (ANN) to SNN conversion for Spiking Transformers that achieves high accuracy, low latency, and low power consumption on complex datasets. The source codes of the proposed method are available at https://github.com/h-z-h-cell/Transformer-to-SNN-ECMT.

Recently, a vast number of image generation models have been proposed, which raises concerns regarding the misuse of these artificial intelligence (AI) techniques for generating fake images. To attribute the AI-generated images, existing schemes usually design and train deep neural networks (DNNs) to learn the model fingerprints, which usually requires a large amount of data for effective learning. In this paper, we aim to answer the following two questions for AI-generated image attribution, 1) is it possible to design useful handcrafted filters to facilitate the fingerprint learning? and 2) how we could reduce the amount of training data after we incorporate the handcrafted filters? We first propose a set of Multi-Directional High-Pass Filters (MHFs) which are capable to extract the subtle fingerprints from various directions. Then, we propose a Directional Enhanced Feature Learning network (DEFL) to take both the MHFs and randomly-initialized filters into consideration. The output of the DEFL is fused with the semantic features to produce a compact fingerprint. To make the compact fingerprint discriminative among different models, we propose a Dual-Margin Contrastive (DMC) loss to tune our DEFL. Finally, we propose a reference based fingerprint classification scheme for image attribution. Experimental results demonstrate that it is indeed helpful to use our MHFs for attributing the AI-generated images. The performance of our proposed method is significantly better than the state-of-the-art for both the closed-set and open-set image attribution, where only a small amount of images are required for training.

Multi-modal Large Language Models (MLLMs) have demonstrated remarkable performance on various visual-language understanding and generation tasks. However, MLLMs occasionally generate content inconsistent with the given images, which is known as "hallucination". Prior works primarily center on evaluating hallucination using standard, unperturbed benchmarks, which overlook the prevalent occurrence of perturbed inputs in real-world scenarios-such as image cropping or blurring-that are critical for a comprehensive assessment of MLLMs' hallucination. In this paper, to bridge this gap, we propose Hallu-PI, the first benchmark designed to evaluate Hallucination in MLLMs within Perturbed Inputs. Specifically, Hallu-PI consists of seven perturbed scenarios, containing 1,260 perturbed images from 11 object types. Each image is accompanied by detailed annotations, which include fine-grained hallucination types, such as existence, attribute, and relation. We equip these annotations with a rich set of questions, making Hallu-PI suitable for both discriminative and generative tasks. Extensive experiments on 12 mainstream MLLMs, such as GPT-4V and Gemini-Pro Vision, demonstrate that these models exhibit significant hallucinations on Hallu-PI, which is not observed in unperturbed scenarios. Furthermore, our research reveals a severe bias in MLLMs' ability to handle different types of hallucinations. We also design two baselines specifically for perturbed scenarios, namely Perturbed-Reminder and Perturbed-ICL. We hope that our study will bring researchers' attention to the limitations of MLLMs when dealing with perturbed inputs, and spur further investigations to address this issue. Our code and datasets are publicly available at https://github.com/NJUNLP/Hallu-PI.

In the field of Vision-Language Models (VLM), the Contrastive Language-Image Pretraining (CLIP) model has yielded outstanding performance on many downstream tasks through prompt tuning. By integrating image and text representations, CLIP exhibits zero-shot generalization capabilities on unseen data. However, when new categories and distribution shifts occur, the pretrained text embeddings in CLIP may not align well with unseen images, potentially leading to a decrease in CLIP's zero-shot generalization performance. To address this issue, many existing methods use test samples to update the CLIP model during testing through a process known as Test-Time Adaptation (TTA). Previous TTA techniques, such as image augmentation, can lead to overfitting given outlying samples, while methods based on teacher-student distillation can increase memory use. Further, these methods significantly increase inference time, which is a crucial factor in the testing phase. To improve robustness, mitigate overfitting, and reduce bias toward outlying samples, we propose a novel method: Self-Text Distillation with Conjugate Pseudo-labels (SCP), designed to enhance CLIP's zero-shot generalization. SCP uses gradient information from conjugate pseudo-labels to enhance the model's robustness toward distribution shifts. It also innovates by using a fixed prompt list to distil learnable prompts from within the same model, acting as a self-regulation mechanism that minimizes overfitting. Additionally, SCP is a fully test-time adaptation method that does not require retraining. It directly improves CLIP's zero-shot generalization at test time without increasing either memory overheads or inference time. In evaluations across three zero-shot generalization scenarios, SCP surpasses existing state-of-the-art methods in performance and significantly reduces inference time.

Human-object interactions (HOI) detection aims at capturing human-object pairs in images and corresponding actions. It is an important step toward high-level visual reasoning and scene understanding. However, due to the natural bias from the real world, existing methods mostly struggle with rare human-object pairs and lead to sub-optimal results. Recently, with the development of the generative model, a straightforward approach is to construct a more balanced dataset based on a group of supplementary samples. Unfortunately, there is a significant domain gap between the generated data and the original data, and simply merging the generated images into the original dataset cannot significantly boost the performance. To alleviate the above problem, we present a novel model-agnostic framework called Context-Enhanced Feature Alignment (CEFA) module, which can effectively align the generated data with the original data at the feature level and bridge the domain gap. Specifically, CEFA consists of a feature alignment module and a context enhancement module. On one hand, considering the crucial role of human-object pairs information in HOI tasks, the feature alignment module aligns the human-object pairs by aggregating instance information. On the other hand, to mitigate the issue of losing important context information caused by the traditional discriminator-style alignment method, we employ a context-enhanced image reconstruction module to improve the model's learning ability of contextual cues. Extensive experiments have shown that our method can serve as a plug-and-play module to improve the detection performance of HOI models on rare categories.

As a critical component in graphic design, artistic posters are widely applied in the advertising and entertainment industry, thus the automatic poster creation from user-provided prompts has become increasingly desired recently. Although existing Text2Image methods create impressive images aligned with given prompts, they fail to generate ideal artistic posters, especially with Chinese texts. To create desired artistic Chinese posters including an aligned background, reasonable layouts, and stylized graphical texts from given prompts only, we propose an automatic poster creation framework, named Prompt2Poster. Our framework utilizes the capacity of the powerful Large Language Model (LLM) to extract user intention from provided prompts and generate the aligned background. Although only taking a user prompt as the input, linguistic, visual, and geometrical information is fully utilized in the framework, bringing the ability to fit different distributions. To achieve the use of multi-modal information in the framework, two carefully designed modules, Controllable Layout Generator (CLG) and Graphical Text Generator (GTG) are proposed, leading to accurate and pleasurable visual results. Comprehensive experiments demonstrate that our Prompt2Poster achieves superior performance, especially in text quality and visual harmony.

Recently, diffusion models have made significant strides in synthesizing realistic 2D human images based on provided text prompts. Building upon this, researchers have extended 2D text-to-image diffusion models into the 3D domain for generating human textures (UV Maps). However, some important problems about UV Map Generative models are still not solved, i.e., how to generate personalized texture maps for any given face image, and how to define and evaluate the quality of these generated texture maps. To solve the above problems, we introduce a novel method, UVMap-ID, which is a controllable and personalized UV Map generative model. Unlike traditional large-scale training methods in 2D, we propose to fine-tune a pre-trained text-to-image diffusion model which is integrated with a face fusion module for achieving ID-driven customized generation. To support the finetuning strategy, we introduce a small-scale attribute-balanced training dataset, including high-quality textures with labeled text and Face ID. Additionally, we introduce some metrics to evaluate the multiple aspects of the textures. Finally, both quantitative and qualitative analyses demonstrate the effectiveness of our method in controllable and personalized UV Map generation.

Large Language Models (LLMs), benefiting from the auto-regressive modelling approach performed on massive unannotated texts corpora, demonstrates powerful perceptual and reasoning capabilities. However, as for extending auto-regressive modelling to multi-modal scenarios to build Large Multi-modal Models (LMMs), there lies a great difficulty that the image information is processed in the LMM as continuous visual embeddings, which cannot obtain discrete supervised labels for classification. In this paper, we successfully perform multi-modal auto-regressive modeling with a unified objective for the first time. Specifically, we propose the concept of visual tokens, which maps the visual features to probability distributions over LLM's vocabulary, providing supervision information for visual modelling. We further explore the distribution of visual features in the semantic space within LMM and the possibility of using text embeddings to represent visual information. Experimental results and ablation studies on 5 VQA tasks and 4 benchmark toolkits validate the powerful performance of our proposed approach.

Due to the limitations of infrared image acquisition conditions, many essential tasks currently rely on visible images as the main source of training data. However, single-modal data makes it difficult for downstream networks to show optimal performance. Therefore, converting the more easily obtainable visible images into infrared images emerges as an effective remedy to alleviate the critical shortage of infrared data. Yet current methods typically focus solely on transferring visible images to infrared style, while overlooking the crucial infrared thermal feature during cross-modal translation. To elevate the authenticity of cross-model translation at the feature level, this paper introduces a translation network based on frequency feature mapping and dual patches contrast, MappingFormer, which can achieve cross-modal image generation from visible to infrared. Specifically, the generator incorporates two branches: low-frequency feature mapping (LFM) and high-frequency feature refinement (HFR), both have embedded the Swin Transformer blocks. The LFM branch captures the fuzzy structural from visible images, while the HFR focuses on mapping edge and texture features. The extracted dual-branch frequency features undergo refinement and fusion through cross-attention mechanisms. Additionally, a dual contrast learning mechanism based on feature patch (DFPC) is designed to infer effective mappings between unaligned cross-modal data. Numerous experimental results prove the effectiveness of this method in cross-modal feature mapping and image generation from visible to infrared. This method holds significant potential for downstream tasks where infrared data is limited.

Music source separation and pitch estimation are two vital tasks in music information retrieval. Typically, the input of pitch estimation is obtained from the output of music source separation. Therefore, existing methods have tried to perform these two tasks simultaneously, so as to leverage the mutually beneficial relationship between both tasks. However, these methods still face two critical challenges that limit the improvement of both tasks: the lack of labeled data and joint learning optimization. To address these challenges, we propose a Model-Agnostic Joint Learning (MAJL) framework for both tasks. MAJL is a generic framework and can use variant models for each task. It includes a two-stage training method and a dynamic weighting method named Dynamic Weights on Hard Samples (DWHS), which addresses the lack of labeled data and joint learning optimization, respectively. Experimental results on public music datasets show that MAJL outperforms state-of-the-art methods on both tasks, with significant improvements of 0.92 in Signal-to-Distortion Ratio (SDR) for music source separation and 2.71% in Raw Pitch Accuracy (RPA) for pitch estimation. Furthermore, comprehensive studies not only validate the effectiveness of each component of MAJL, but also indicate the great generality of MAJL in adapting to different model architectures.

Multi-View Clustering (MVC) aims to mine complementary information across different views to partition multi-view data more effectively and has attracted considerable interest. However, existing deep multi-view clustering methods frequently neglect the exploration of structural information within individual view and lack the learning of structural consistency among views, which results in limitations in the clustering performance. In this paper, we introduce a novel multi-view clustering framework based on graph consistency learning to address this issue. Specifically, we design intra-view graph contrastive learning to uncover structural information within each view and achieve structural conscistency objectives through cross-view graph consistency learning. Additionally, to address the conflict between different learning objectives when trained in the same space, we introduce two new feature spaces, one for cluster-levcel contrastive learning and the other for instance-level contrastive learning. Subsequently, to make the most of discriminative information from all views, we concatenate high-level features from all views to form global features and employ self-supervision to promote clustering consistency across different views. Experimental results on several challenging datasets demonstrate the outstanding performance of our proposed method.

Due to device constraints and lighting conditions, captured images frequently exhibit coupled low-resolution and ultra-dark degradations. Enhancing the visibility and resolution of ultra-dark images simultaneously is crucial for practical applications. Current approaches often address both tasks in isolation or through simplistic cascading strategies, while also relying heavily on empirical and manually designed composite loss constraints, which inevitably results in compromised training efficacy, increased artifacts, and diminished detail fidelity. To address these issues, we propose TriCo, the first to adopt a Tri -level learning framework that explicitly formulates the bidirectional Co operative relationship and devises algorithms to tackle coupled degradation factors. In the optimization across Upper (U)-Middle (M)-Lower (L) levels, we model the synergistic dependencies between illumination learning and super-resolution tasks within the M-L levels. Moving to the U-M levels, we introduce hyper-variables to automate the learning of beneficial constraints for both learning tasks, moving beyond the traditional trial-and-error pitfalls of the learning process. Algorithmically, we establish a Phased Gradient-Response (PGR) algorithm as our training mechanism, which facilitates a dynamic, inter-variable gradient feedback and ensures efficient and rapid convergence. Moreover, we merge inherent illumination priors with universal semantic model features to adaptively guide pixel-level high-frequency detail recovery. Extensive experimentation validates the framework's broad generalizability across challenging ultra-dark scenarios, outperforming current state-of-the-art methods across 4 real and synthetic benchmark datasets over 6 metrics (e.g., 5.8%← in PSNR and 26.6%← in LPIPS).

The rapid development of multi-media techniques boosts the emergence of multi-view data, and how to uncover its intrinsic structure and utilize it to conduct the subsequent downstream tasks is crucial in data analysis. Multi-view clustering is representative of handling multi-view data. The anchor-based method has received widespread attention for excellent performance and low time complexity. However, existing methods encounter two drawbacks, cutting down their performance, i.e., the assumption of the availability of all views and limited interaction of anchor generation among views. In some scenes, views arrive sequentially, and storing them is challenging owing to the limited space/privacy considerations, and the existing anchor-based MVC is unsuitable for this. Additionally, recent works fail to generate anchors with the guidance of other views, and it is tough to align the anchor graphs. To this end, we propose A Lightweight Anchor-Based Incremental Framework for Multi-view Clustering. Specifically, we first initialize an anchor graph with the assistance of k-means when a new view arrives. Then, the consensus one of the anchor graph is updated by the newly collected view with a permutation matrix. Our proposed method is more capable of anchor alignment because, in incremental MVC, the anchor graphs of previous views could be listed as a reference to guide the generation of anchor graphs of the coming view. Furthermore, we design a three-step iterative and convergent algorithm to address the resultant problem. Notably, the proposed algorithm shows outstanding effectiveness and time/space efficiency in extensive experiments.

Image-to-image translation is defined as the process of learning a mapping between images from a source domain and images from a target domain. The probabilistic structure that maps a fixed initial state to a pinned terminal state through a standard Wiener process is a Brownian bridge. In this paper, we propose a score-based Stochastic Differential Equation (SDE) approach via the Brownian bridges, termed the Amenable Brownian Bridges (A-Bridges), to image-to-image translation tasks as an unconditional diffusion model. Our framework embraces a large family of Brownian bridge models, while the discretization of the linear A-Bridge exploits its advantage that provides the explicit solution in a closed form and thus facilitates the model training. Our model enables the accelerated sampling and has achieved record-breaking performance in sample quality and diversity on benchmark datasets following the guidance of its SDE structure.