ACM MM 2024 论文列表

The latest Text-to-Speech (TTS) systems can produce speech with voice quality and naturalness comparable to human speech. Yet the demand for large amount of high-quality data from target speakers remains a significant challenge. Particularly for long-form expressive reading, target speaker's training speech that covers rich contextual information are needed. In this paper a novel design of context-aware speech pre-trained model is developed for expressive TTS based on contrastive learning. The model can be trained with abundant speech data without explicitly labelled speaker identities. It captures the intricate relationship between the speech expression of a spoken sentence and the contextual text information. By incorporating cross-modal text and speech features into the TTS model, it enables the generation of coherent and expressive speech, which is especially beneficial when there is a scarcity of target speaker data. The pre-trained model is evaluated first in the task of Context-Speech retrieval and then as the integral part of a zero-shot TTS system. Experimental results demonstrate that the pretraining framework effectively learns Context-Speech representations and significantly enhances the expressiveness of synthesized speech. Audio demos are available at: https://ccsp2024.github.io/demo/.

Multi-view based molecular properties prediction learning has received widely attention in recent years in terms of its potential for the downstream tasks in the field of drug discovery. However, the consistency of different molecular view representations and the full utilization of complementary information among them in existing multi-view molecular property prediction methods remain to be further explored. Furthermore, most current methods focus on generating global level representations at the graph level with information from different molecular views (e.g., 2D and 3D views) assuming that the information can be corresponded to each other. In fact it is not unusual that for example the conformation change or computational errors may lead to discrepancies between views. To addressing these issues, we propose a new Cross-View contrastive unification guides Generative Molcular pre-trained model, call MolCVG. We first focus on common and private information extraction from 2D graph views and 3D geometric views of molecules, Minimizing the impact of noise in private information on subsequent strategies. To exploit both types of information in a more refined way, we propose a cross-view contrastive unification strategy to learn cross-view global information and guide the reconstruction of masked nodes, thus effectively optimizing global features and local descriptions. Extensive experiments on real-world molecular data sets demonstrate the effectiveness of our approach for molecular property prediction task.

The Text to Audible-Video Generation (TAVG) task involves generating videos with accompanying audio based on text descriptions. Achieving this requires skillful alignment of both audio and video elements. To support research in this field, we have developed a comprehensive Text to Audible-Video Generation Benchmark (TAVGBench), which contains over 1.7 million clips with a total duration of 11.8 thousand hours. We propose an automatic annotation pipeline to ensure each audible video has detailed descriptions for both its audio and video contents. We also introduce the Audio-Visual Harmoni score (AVHScore) to provide a quantitative measure of the alignment between the generated audio and video modalities. Additionally, we present a baseline model for TAVG called TAVDiffusion, which uses a two-stream latent diffusion model to provide a fundamental starting point for further research in this area. We achieve the alignment of audio and video by employing cross-attention and contrastive learning. Through extensive experiments and evaluations on TAVGBench, we demonstrate the effectiveness of our proposed model under both conventional metrics and our proposed metrics. The dataset and code can be found on this page https://npucvr.github.io/TAVGBench/ and on github https://github.com/OpenNLPLab/TAVGBench.

With the growth of VR and AR industry, 3D reconstruction has become a more and more important topic in multimedia. Although 3D Gaussian Splatting achieves state-of-the-art in 3D Reconstruction, a large number of Gaussians are needed to fit a 3D scene due to the Gibbs Phenomenon. The pursuit of compressing 3D Gaussian Splatting and reducing memory overhead has long been a focal point. Embarking on this trajectory, our study delves into this domain, aiming to mitigate these challenges. Inspired by the tangram, a Chinese ancient puzzle, we introduce a novel methodology (Tangram-Splatting) that leverages shape priors to optimize 3D scene fitting. Central to our approach is a pioneering technique that diversifies Gaussian function types while preserving algorithmic efficiency. Through exhaustive experimentation, we demonstrate that our method achieves a remarkable average reduction of 62.4% in memory consumption used to store optimized parameters and decreases the training time by at least 10 minutes, with only marginal sacrifices in PSNR performance, typically under 0.3 dB, and our algorithm is even better on some datasets. This reduction in memory burden is of paramount significance for real-world applications, mitigating the substantial memory footprint and transmission burden traditionally associated with such algorithms. Our algorithm underscores the profound potential of Tangram-Splatting in advancing multimedia applications.

Continual learning (CL) breaks off the one-way training manner and enables a model to adapt to new data, semantics and tasks continuously. However, current CL methods mainly focus on single tasks. Besides, CL models are plagued by catastrophic forgetting and semantic drift since the lack of old data, which often occurs in remote-sensing interpretation due to the intricate fine-grained semantics. In this paper, we propose Continual Panoptic Perception (CPP), a unified continual learning model that leverages multi-task joint learning covering pixel-level classification, instance-level segmentation and image-level perception for universal interpretation in remote sensing images. Concretely, we propose a collaborative cross-modal encoder (CCE) to extract the input image features, which supports pixel classification and caption generation synchronously. To inherit the knowledge from the old model without exemplar memory, we propose a task-interactive knowledge distillation (TKD) method, which leverages cross-modal optimization and task-asymmetric pseudo-labeling (TPL) to alleviate catastrophic forgetting. Furthermore, we also propose a joint optimization mechanism to achieve end-to-end multi-modal panoptic perception. Experimental results on the fine-grained panoptic perception dataset validate the effectiveness of the proposed model, and also prove that joint optimization can boost sub-task CL efficiency with over 13% relative improvement on panoptic quality. The project page is available at https://github.com/YBIO/CPP.

With the increasing popularity of online social applications, stickers have become common in online chats. Teaching a model to select the appropriate sticker from a set of candidate stickers based on dialogue context is important for optimizing the user experience. Existing methods have proposed leveraging emotional information to facilitate the selection of appropriate stickers. However, considering the frequent co-occurrence among sticker images, words with emotional preference in the dialogue and emotion labels, these methods tend to over-rely on such dataset bias, inducing spurious correlations during training. As a result, these methods may select inappropriate stickers that do not match users' intended expression. In this paper, we introduce a causal graph to explicitly identify the spurious correlations in the sticker selection task. Building upon the analysis, we propose a Causal Knowledge-Enhanced Sticker Selection model to mitigate spurious correlations. Specifically, we design a knowledge-enhanced emotional utterance extractor to identify emotional information within dialogues. Then an interventional visual feature extractor is employed to obtain unbiased visual features, aligning them with the emotional utterances representation. Finally, a standard transformer encoder fuses the multimodal information for emotion recognition and sticker selection. Extensive experiments on the MOD dataset show that our CKS model significantly outperforms the baseline models.

Large 2D vision-language models (2D-LLMs) have gained significant attention by bridging Large Language Models (LLMs) with images using a simple projector. Inspired by their success, large 3D point cloud-language models (3D-LLMs) also integrate point clouds into LLMs. However, directly aligning point clouds with LLM requires expensive training costs, typically in hundreds of GPU-hours on A100, which hinders the development of 3D-LLMs. In this paper, we introduce MiniGPT-3D, an efficient and powerful 3D-LLM that achieves multiple SOTA results while training for only 27 hours on one RTX 3090. Specifically, we propose to align 3D point clouds with LLMs using 2D priors from 2D-LLMs, which can leverage the similarity between 2D and 3D visual information. We introduce a novel four-stage training strategy for modality alignment in a cascaded way, and a mixture of query experts module to adaptively aggregate features with high efficiency. Moreover, we utilize parameter-efficient fine-tuning methods LoRA and Norm fine-tuning, resulting in only 47.8M learnable parameters, which is up to 260x fewer than existing methods. Extensive experiments show that MiniGPT-3D achieves SOTA on 3D object classification and captioning tasks, with significantly cheaper training costs. Notably, MiniGPT-3D gains an 8.12 increase on GPT-4 evaluation score for the challenging object captioning task compared to ShapeLLM-13B, while the latter costs 160 total GPU-hours on 8 A800. We are the first to explore the efficient 3D-LLM, offering new insights to the community. Code and weights are available at https://github.com/TangYuan96/MiniGPT-3D.

Although deep learning-based methods have made significant advances in the field of image restoration (IR), they often suffer from excessive model parameters. To tackle this problem, this work proposes a compact Transformer (Compacter) for lightweight image restoration by making several key designs. We employ the concepts of projection sharing, adaptive interaction, and heterogeneous aggregation to develop a novel Compact Adaptive Self-Attention (CASA). Specifically, CASA utilizes shared projection to generate Query, Key, and Value to simultaneously model spatial and channel-wise self-attention. The adaptive interaction process is then used to propagate and integrate global information from two different dimensions, thus enabling omnidirectional relational interaction. Finally, a depth-wise convolution is incorporated on Value to complement heterogeneous local information, enabling global-local coupling. Moreover, we propose a Dual Selective Gated Module (DSGM) to dynamically encapsulate the globality into each pixel for context-adaptive aggregation. Extensive experiments demonstrate that our Compacter achieves state-of-the-art performance for a variety of lightweight IR tasks with approximately 400K parameters.

As machine learning advances, machine learning as a service (MLaaS) in the cloud brings convenience to human lives but also privacy risks, as powerful neural networks used for generation, classification or other tasks can also become privacy snoopers. This motivates privacy preservation in the inference phase. Many approaches for preserving privacy in the inference phase introduce multi-objective functions, training models to remove specific private information from users' uploaded data. Although effective, these adversarial learning-based approaches suffer not only from convergence difficulties, but also from limited generalization beyond the specific privacy for which they are trained. To address these issues, we propose a method for privacy preservation in the inference phase by removing task-irrelevant information, which requires no knowledge of the privacy attacks nor introduction of adversarial learning. Specifically, we introduce a metric to distinguish task-irrelevant information from task-relevant information, and achieve more efficient metric estimation to remove task-irrelevant features. The experiments demonstrate the potential of our method in several tasks. Our code will be available at: https://github.com/iwhoyoung/PriFU.

Accurate segmentation of cerebrovascular structures from Time-of-flight magnetic resonance angiography is vital for treating cerebrovascular diseases. However, existing methods rely on voxel categorization, leading to discontinuities in fine vessel locations. We propose a connectivity-based cerebrovascular segmentation method that considers inter-voxel relationships to alleviate this limitation. By modeling connectivity, we convert voxel classification into inter-voxel connectivity prediction. Given the sparse and widely distributed nature of cerebrovascular structures, we employ a sparse 3D Bi-level routing attention to effectively capture cerebrovascular features. To extract inter-voxel directional information, we utilize a 3D direction excitation block. Additionally, a 3D direction interactive block continuously enhances the directional information within the feature map. This directional enhanced feature map is then concatenated with the feature map from the encoder layer and fed into the decoder layer. We compare our method with current state-of-the-art cerebrovascular segmentation techniques and general medical image segmentation methods. Our method achieved a Dice Similarity Coefficient of 92.413% and 83.197% on the clinical and open cerebrovascular datasets respectively, outperforming existing approaches.

When applying high-level visual algorithms to rainy scenes, it is customary to preprocess the rainy images using low-level rain removal networks, followed by visual networks to achieve the desired objectives. Such a setting has never been explored by adversarial attack methods, which are only limited to attacking one kind of them. Considering the deficiency of multi-functional attacking strategies and the significance for open-world perception scenarios, we are the first to propose a Cascaded Adversarial Attack (CAA) setting, where the adversarial example can simultaneously attack different-level tasks, such as rain removal and semantic segmentation in an integrated system. Specifically, our attack on the rain removal network aims to preserve rain streaks in the output image, while for the semantic segmentation network, we employ powerful existing adversarial attack methods to induce misclassification of the image content. Importantly, CAA innovatively utilizes binary masks to effectively concentrate the aforementioned two significantly disparate perturbation distributions on the input image, enabling attacks on both networks. Additionally, we propose two variants of CAA, which minimize the differences between the two generated perturbations by introducing a carefully designed perturbation interaction mechanism, resulting in enhanced attack performance. Extensive experiments validate the effectiveness of our methods, demonstrating their superior ability to significantly degrade the performance of the downstream task compared to methods that solely attack a single network.

The modeling and manipulation of 3D scenes captured from the real world are pivotal in various applications, attracting growing research interest. While previous works on editing have achieved interesting results through manipulating 3D meshes, they often require accurately reconstructed meshes to perform editing, which limits their application in 3D content generation. To address this gap, we introduce a novel single-image-driven 3D scene editing approach based on 3D Gaussian Splatting, enabling intuitive manipulation via directly editing the content on a 2D image plane. Our method learns to optimize the 3D Gaussians to align with an edited version of the image rendered from a user-specified viewpoint of the original scene. To capture long-range object deformation, we introduce positional loss into the optimization process of 3D Gaussian Splatting and enable gradient propagation through reparameterization. To handle occluded 3D Gaussians when rendering from the specified viewpoint, we build an anchor-based structure and employ a coarse-to-fine optimization strategy capable of handling long-range deformation while maintaining structural stability. Furthermore, we design a novel masking strategy to adaptively identify non-rigid deformation regions for fine-scale modeling. Extensive experiments show the effectiveness of our method in handling geometric details, long-range, and non-rigid deformation, demonstrating superior editing flexibility and quality compared to previous approaches.

Real-world data consistently exhibits a long-tailed distribution, often spanning multiple categories. This complexity underscores the challenge of content comprehension, particularly in scenarios requiring Long-Tailed Multi-Label image Classification (LTMLC). In such contexts, imbalanced data distribution and multi-object recognition pose significant hurdles. To address this issue, we propose a novel and effective approach for LTMLC, termed Category-Prompt Refined Feature Learning (CPRFL), utilizing semantic correlations between different categories and decoupling category-specific visual representations for each category. Specifically, CPRFL initializes category-prompts from the pretrained CLIP's embeddings and decouples category-specific visual representations through interaction with visual features, thereby facilitating the establishment of semantic correlations between the head and tail classes. To mitigate the visual-semantic domain bias, we design a progressive Dual-Path Back-Propagation mechanism to refine the prompts by progressively incorporating context-related visual information into prompts. Simultaneously, the refinement process facilitates the progressive purification of the category-specific visual representations under the guidance of the refined prompts. Furthermore, taking into account the negative-positive sample imbalance, we adopt the Asymmetric Loss as our optimization objective to suppress negative samples across all classes and potentially enhance the head-to-tail recognition performance. We validate the effectiveness of our method on two LTMLC benchmarks and extensive experiments demonstrate the superiority of our work over baselines.The code is available at https://github.com/jiexuanyan/CPRFL.

In the realm of Medical Visual Language Models (Med-VLMs), the quest for universal efficient fine-tuning mechanisms remains paramount, especially given researchers in interdisciplinary fields are often extremely short of training resources, yet largely unexplored. Given the unique challenges in the medical domain, such as limited data scope and significant domain-specific requirements, evaluating and adapting Parameter-Efficient Fine-Tuning (PEFT) methods specifically for Med-VLMs is essential. Most of the current PEFT methods on Med-VLMs have yet to be comprehensively investigated but mainly focus on adding some components to the model's structure or input. However, fine-tuning intrinsic model components often yields better generality and consistency, and its impact on the ultimate performance of Med-VLMs has been widely overlooked and remains understudied. In this paper, we endeavour to explore an alternative to traditional PEFT methods, especially the impact of fine-tuning LayerNorm layers, FFNs and Attention layers on the Med-VLMs. Our comprehensive studies span both small-scale and large-scale Med-VLMs, evaluating their performance under various fine-tuning paradigms across tasks such as Medical Visual Question Answering and Medical Imaging Report Generation. The findings reveal unique insights into the effects of intrinsic parameter fine-tuning methods on fine-tuning Med-VLMs to downstream tasks and expose fine-tuning solely the LayerNorm layers not only surpasses the efficiency of traditional PEFT methods but also retains the model's accuracy and generalization capabilities across a spectrum of medical downstream tasks. The experiments show LayerNorm fine-tuning's superior adaptability and scalability, particularly in the context of large-scale Med-VLMs.

3D multimodal question answering (MQA) plays a crucial role in scene understanding by enabling intelligent agents to comprehend their surroundings in 3D environments. While existing research has primarily focused on indoor household tasks and outdoor roadside autonomous driving tasks, there has been limited exploration of city-level scene understanding tasks. Furthermore, existing research faces challenges in understanding city scenes, due to the absence of spatial semantic information and human-environment interaction information at the city level.To address these challenges, we investigate 3D MQA from both dataset and method perspectives. From the dataset perspective, we introduce a novel 3D MQA dataset named City-3DQA for city-level scene understanding, which is the first dataset to incorporate scene semantic and human-environment interactive tasks within the city. From the method perspective, we propose a Scene graph enhanced City-level Understanding method (Sg-CityU), which utilizes the scene graph to introduce the spatial semantic. A new benchmark is reported and our proposed Sg-CityU achieves accuracy of 63.94 % and 63.76 % in different settings of City-3DQA. Compared to indoor 3D MQA methods and zero-shot using advanced large language models (LLMs), Sg-CityU demonstrates state-of-the-art (SOTA) performance in robustness and generalization.

Adversarial attacks on point clouds are crucial for assessing and improving the adversarial robustness of 3D deep learning models. Despite leveraging various geometric constraints, current adversarial attack strategies often suffer from inadequate imperceptibility. Given that adversarial perturbations tend to disrupt the inherent symmetry in objects, we recognize this disruption as the primary cause of the lack of imperceptibility in these attacks. In this paper, we introduce a novel framework, symmetry-aware imperceptible adversarial attacks on 3D point clouds (SymAttack), to address this issue. Our approach starts by identifying part- and patch-level symmetry elements, and grouping points based on semantic and Euclidean distances, respectively. During the adversarial attack iterations, we intentionally adjust the perturbation vectors on symmetric points relative to their symmetry plane. By preserving symmetry within the attack process, SymAttack significantly enhances imperceptibility. Extensive experiments validate the effectiveness of SymAttack in generating imperceptible adversarial point clouds, demonstrating its superiority over the state-of-the-art methods.

Diffusion models have shown remarkable prowess in text-to-image synthesis and editing, yet they often stumble when tasked with interpreting complex prompts that describe multiple entities with specific attributes and interrelations. The generated images often contain inconsistent multi-entity representation (IMR), reflected as inaccurate presentations of the multiple entities and their attributes. Although providing spatial layout guidance improves the multientity generation quality in existing works, it is still challenging to handle the leakage attributes and avoid unnatural characteristics. To address the IMR challenge, we first conduct in-depth analyses of the diffusion process and attention operation, revealing that the IMR challenges largely stem from the process of cross-attention mechanisms. According to the analyses, we introduce the entity guidance generation mechanism, which maintains the integrity of the original diffusion model parameters by integrating plugin networks. Our work advances the stable diffusion model by segmenting comprehensive prompts into distinct entity-specific prompts with bounding boxes, enabling a transition from multientity to single-entity generation in cross-attention layers. More importantly, we introduce entity-centric cross-attention layers that focus on individual entities to preserve their uniqueness and accuracy, alongside global entity alignment layers that refine crossattention maps using multi-entity priors for precise positioning and attribute accuracy. Additionally, a linear attenuation module is integrated to progressively reduce the influence of these layers during inference, preventing oversaturation and preserving generation fidelity. Our comprehensive experiments demonstrate that this entity guidance generation enhances existing text-to-image models in generating detailed, multi-entity images. Code is available at https://github.com/chaos-sun/eggen.git.

The issue of face privacy protection has aroused wide social concern along with the increasing applications of face images. The latest methods focus on achieving a good privacy-utility tradeoff so that the protected results can still be used to support the downstream computer vision tasks. However, they may suffer from limited flexibility in manipulating this tradeoff because the practical requirements may vary under different scenarios. In this paper, we present a novel recurrent latent representation reorganization (LReOrg) framework to deal with the problem. LReOrg relies on two key modules to deal with the privacy-utility tradeoff, where the first one is responsible for anonymizing the privacy sensitive information and the other is responsible for recovering the destroyed useful insensitive information according to user requirements. LReOrg is advantageous in: (a) enabling users to recurrently process fine-grained attributes; (b) providing flexible control over privacy-utility tradeoff by manipulating which attributes to anonymize or preserve using cross-modal keywords; and (c) eliminating the need of data annotations for network training. The experimental results on benchmark datasets have reported the superior ability of our approach for providing flexible protection on facial information.