ACM MM 2024 论文列表

Singing melody extraction is a key task in the field of music information retrieval (MIR). However, decades of research works have uncovered two difficult issues. First, binary classification on frequency-domain audio features (e.g., spectrogram) is regarded as the primary method, which ignores the potential associations of musical information at different frequency bins, as well as their varying significance for output decisions. Second, the existing semi-supervised singing melody extraction models ignore the accuracy of the generated pseudo labels by semi-supervised models, which largely limits the further improvements of the model. To solve the two issues, in this paper, we propose a heterogeneous knowledge distillation framework for semi-supervised singing melody extraction using harmonic supervision, termed as HKDSME. We begin by proposing a four-class classification paradigm for determining the results of singing melody extraction using harmonic supervision. This enables the model to capture more information regarding melodic relations in spectrograms. To improve the accuracy issue of pseudo labels, we then build a semi-supervised method by leveraging the extracted harmonics as a consistent regularization. Different from previous methods, it judges the availability of unlabeled data in terms of the inner positional relations of extracted harmonics. To further build a light-weight semi-supervised model, we propose a heterogeneous knowledge distillation (HKD) module, which enables the prior knowledge to transfer between heterogeneous models. We also propose a novel confidence guided loss, which incorporates with the proposed HKD module to reduce the wrong pseudo labels. We evaluate our proposed method using several well-known public available datasets, and the findings demonstrate the efficacy of our proposed method.

In real-world photography, local motion blur often arises from the interplay between moving objects and stationary backgrounds during exposure. Existing deblurring methods face challenges in addressing local motion deblurring due to (i) the presence of arbitrary localized blurs and uncertain blur extents; (ii) the limited ability to accurately identify specific blurs resulting from ambiguous motion boundaries. These limitations often lead to suboptimal solutions when estimating blur maps and generating final deblurred images. To that end, we propose a novel method named Motion-Uncertainty-Guided Network (MUGNet), which harnesses a probabilistic representational model to explicitly address the intricacies stemming from motion uncertainties. Specifically, MUGNet consists of two key components, i.e., motion-uncertainty quantification (MUQ) module and motion-masked separable attention (M2SA) module, serving for complementary purposes. Concretely, MUQ aims to learn a conditional distribution for accurate and reliable blur map estimation, while the M2SA module is to enhance the representation of regions influenced by local motion blur and static background, which is achieved by promoting the establishment of extensive global interactions. We demonstrate the superiority of our MUGNet with extensive experiments. The code is publicly available at: https://github.com/zeyuxiao1997/MUGNet.

QUIC is the underlying protocol of the next generation HTTP/3, serving as the major vehicle delivering video data nowadays. As a userspace protocol based on UDP, QUIC features low transmission latency and has been widely deployed by content providers. However, the high computational overhead of QUIC shifts system knobs to CPUs in high-bandwidth scenarios. When CPU resources become the constraint, HTTP/3 exhibits even lower throughput than HTTP/1.1. In this paper, we carefully analyze the performance bottleneck of QUIC and find it results from ACK processing, packet sending, and data encryption. By reducing the ACK frequency, activating UDP generic segmentation offload (GSO), and incorporating PicoTLS, a high-performance encryption library, the CPU overhead of QUIC could be effectively reduced in stable network environments. However, simply reducing the ACK frequency also impairs the transmission throughput of QUIC under poor network conditions. To solve this, we develop LiteQUIC, which involves two mechanisms towards alleviating the overhead of ACK processing in addition to GSO and PicoTLS. We evaluate LiteQUIC in the DASH-based video streaming, and the results show that LiteQUIC achieves 1.2× higher average bitrate and 93.3% lower rebuffering time than an optimized version of QUIC with GSO and PicoTLS.

Referring Remote Sensing Image Segmentation (RRSIS) is a challenging task that aims to identify specific regions in aerial images that are relevant to given textual conditions. Existing methods tend to adopt the paradigm of implicit optimization, utilizing a framework consisting of early cross-modal feature fusion and a fixed convolutional kernel-based predictor, neglecting the inherent inter-domain gap and conducting class-agnostic predictions. In this paper, we rethink the issues with the implicit optimization paradigm and address the RRSIS task from a dual-alignment perspective. Specifically, we prepend the dedicated Dual Alignment Network (DANet), including an explicit alignment strategy and a reliable agent alignment module. The explicit alignment strategy effectively reduces domain discrepancies by narrowing the inter-domain affinity distribution. Meanwhile, the reliable agent alignment module aims to enhance the predictor's multi-modality awareness and alleviate the impact of deceptive noise interference. Extensive experiments on two remote sensing datasets demonstrate the effectiveness of our proposed DANet in achieving superior segmentation performance without introducing additional learnable parameters compared to state-of-the-art methods.

Zero-shot anomaly detection (ZSAD) methods detect anomalies without prior access to known normal or abnormal samples within target categories. Existing methods typically rely on pretrained multimodal models, computing similarities between manually crafted textual features representing ''normal'' or ''abnormal'' semantics and image patch features to detect anomalies. However, the generic descriptions of ''abnormal'' often fail to precisely match diverse types of anomalies across different object categories. Additionally, computing feature similarities for single patches struggles to pinpoint specific locations of anomalies with various sizes and scales. To address these issues, we propose a novel ZSAD method called FiLo, comprising two components: adaptively learned Fine-Grained Description (FG-Des) and position-enhanced High-Quality Localization (HQ-Loc). FG-Des introduces fine-grained anomaly descriptions for each category using Large Language Models (LLMs) and employs adaptively learned textual templates to enhance the accuracy and interpretability of anomaly detection. HQ-Loc, utilizing Grounding DINO for preliminary localization, position-enhanced text prompts, and Multi-scale Multi-shape Cross-modal Interaction (MMCI) module, facilitates more accurate localization of anomalies of different sizes and shapes. Experimental results on datasets like MVTec and VisA demonstrate that FiLo significantly improves the performance of ZSAD in both detection and localization, achieving state-of-the-art performance with an image-level AUC of 83.9% and a pixel-level AUC of 95.9% on the VisA dataset. Code is available at https://github.com/CASIA-IVA-Lab/FiLo.

Text-to-image (T2I) generation is a pivotal and core interest within the realm of AI content generation. Amid the swift advancements of both open-source (such as Stable Diffusion) and proprietary (for example, DALLE, MidJourney) T2I models, there is a notable absence of a comprehensive and robust quantitative framework for evaluating their output quality. Traditional methods of quality assessment overlook the textual prompts when judging images; meanwhile, the advent of large multi-modal models (LMMs) introduces the capability to incorporate text prompts in evaluations, yet the challenge of fine-tuning these models for precise T2I quality assessment remains unresolved. In our study, we introduce the T2I-Scorer, a novel two-stage training methodology aimed at fine-tuning LMMs for T2I evaluation. For the first stage, we collect 397K GPT-4V-labeled question-answer pairs related to T2I evaluation. Termed as T2I-ITD, the pseudo-labeled dataset is analyzed and examined by human, and used for instruction tuning to improve the LMM's low-level quality perception. The first stage model, T2I-Scorer-IT, has reached superior accuracy on T2I evaluation than all kinds of existing T2I metrics under zero-shot settings. For the second stage, we define an explicit multi-task training scheme to further align the LMM with human opinion scores, and the fine-tuned T2I-Scorer can reach state-of-the-art accuracy on both image quality and image-text alignment perspectives with significant improvements. We anticipate the proposed metrics can serve as a reliable metric to gauge the ability of T2I generation models in the future. We will make code, data, and weights publicly available.

Drone-based crowd tracking faces difficulties in accurately identifying and monitoring objects from an aerial perspective, largely due to their small size and close proximity to each other, which complicates both localization and tracking. To address these challenges, we present the Density-aware Tracking (DenseTrack) framework. DenseTrack capitalizes on crowd counting to precisely determine object locations, blending visual and motion cues to improve the tracking of small-scale objects. It specifically addresses the problem of cross-frame motion to enhance tracking accuracy and dependability. DenseTrack employs crowd density estimates as anchors for exact object localization within video frames. These estimates are merged with motion and position information from the tracking network, with motion offsets serving as key tracking cues. Moreover, DenseTrack enhances the ability to distinguish small-scale objects using insights from the visual-language model, integrating appearance with motion cues. The framework utilizes the Hungarian algorithm to ensure the accurate matching of individuals across frames. Demonstrated on DroneCrowd dataset, our approach exhibits superior performance, confirming its effectiveness in scenarios captured by drones. Our code will be available at: https://github.com/Zebrabeast/DenseTrack.

Multi-task dense prediction plays an important role in the field of computer vision and has an abundant array of applications. Its main purpose is to reduce the amount of network training parameters by sharing network parameters while using the correlation between tasks to improve overall performance. We propose a task-conditional network that handles one task at a time and shares most network parameters to achieve these goals. Inspired by adapter tuning, we propose an adapter module that focuses on both spatial- and channel-wise information to extract features from the frozen encoder backbone. This approach not only reduces the number of training parameters, but also saves training time and memory resources by attaching a parallel adapter pathway to the encoder. We additionally use learnable task prompts to model different tasks and use these prompts to adjust some parameters of adapters to fit the network to diverse tasks. These task-conditional adapters are also applied to the decoder, which enables the entire network to switch between various tasks, producing better task-specific features and achieving excellent performance. Extensive experiments on two challenging multi-task benchmarks, NYUD-v2 and PASCAL-Context, show that our approach achieves state-of-the-art performance with excellent parameter, time, and memory efficiency. The code is available at https://github.com/jfzleo/Task-Conditional-Adapter.

We introduce HmPEAR, a novel dataset crafted for advancing research in 3D Human Pose Estimation (3D HPE) and Human Action Recognition (HAR), with a primary focus on outdoor environments. This dataset offers a synchronized collection of imagery, LiDAR point clouds, 3D human poses, and action categories. In total, the dataset encompasses over 300,000 frames collected from 10 distinct scenes and 25 diverse subjects. Among these, 250,000 frames of data contain 3D human pose annotations captured using an advanced motion capture system and further optimized for accuracy. Furthermore, the dataset annotates 40 types of daily human actions, resulting in over 6,000 action clips. Through extensive experimentation, we have demonstrated the quality of HmPEAR and highlighted the challenges it presents to current methodologies. Additionally, we propose baselines leveraging sequential images and point clouds for 3D HPE and HAR, which underscore the mutual reinforcement between them, highlighting the potential for cross-task synergies. The dataset is available at http://www.lidarhumanmotion.net/hmpear.

Although large multi-modality models (LMMs) have seen extensive exploration and application in various quality assessment studies, their integration into Point Cloud Quality Assessment (PCQA) remains unexplored. Given LMMs' exceptional performance and robustness in low-level vision and quality assessment tasks, this study aims to investigate the feasibility of imparting PCQA knowledge to LMMs through text supervision. To achieve this, we transform quality labels into textual descriptions during the fine-tuning phase, enabling LMMs to derive quality rating logits from 2D projections of point clouds. To compensate for the loss of perception in the 3D domain, structural features are extracted as well. These quality logits and structural features are then combined and regressed into quality scores. Our experimental results affirm the effectiveness of our approach, showcasing a novel integration of LMMs into PCQA that enhances model understanding and assessment accuracy. We hope our contributions can inspire subsequent investigations into the fusion of LMMs with PCQA, fostering advancements in 3D visual quality analysis and beyond. The code is available at https://github.com/zzc-1998/LMM-PCQA.

Open-domain multi-modal dialogue system heavily relies on visual information to generate contextually relevant responses. The existing open-domain multi-modal dialog generation methods ignore the complementary relationship between multiple modalities, and are difficult to integrate with LLMs. To tackle these challenges, we introduce AutoGraph, an innovative method for constructing visual context graphs automatically. We aim to structure complex information and seamlessly integrate it with large language models (LLMs), aligning information from multiple modalities at both semantic and structural levels. Specifically, we fully connect the text graphs and scene graphs, and then trim unnecessary edges via LLMs to automatically construct a visual context graph. Next, we design several graph sampling grammar for the first time to convert graph structures into sequence which is suitable for LLMs. Finally, we propose a two-stage fine-tuning strategy to allow LLMs to understand graph sampling grammar and generate responses. We validate our proposed method on text-based LLMs, and visual-based LLMs, respectively. Experimental results show that our proposed method achieves state-of-the-art performance on multiple public datasets.

Graph Neural Networks have demonstrated great success in various fields of multimedia. However, the distribution shift between the training and test data challenges the effectiveness of GNNs. To mitigate this challenge, Test-Time Training (TTT) has been proposed as a promising approach. Traditional TTT methods require a demanding unsupervised training strategy to capture the information from test to benefit the main task. Inspired by the great annotation ability of Large Language Models (LLMs) on Text-Attributed Graphs (TAGs), we propose to enhance the test-time training on graphs with LLMs as annotators. In this paper, we design a novel Test-Time Training pipeline, LLMTTT, which conducts the test-time adaptation under the annotations by LLMs on a carefully-selected node set. Specifically, LLMTTT introduces a hybrid active node selection strategy that considers not only node diversity and representativeness, but also prediction signals from the pre-trained model. Given annotations from LLMs, a two-stage training strategy is designed to tailor the test-time model with the limited and noisy labels. A theoretical analysis ensures the validity of our method and extensive experiments demonstrate that the proposed LLMTTT can achieve a significant performance improvement compared to existing Out-of-Distribution (OOD) generalization methods.

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/.

Recent AIGC systems possess the capability to generate digital multimedia content based on human language instructions, such as text, image and video. However, when it comes to speech, existing methods related to human instruction-to-speech generation exhibit two limitations. Firstly, they require the division of inputs into content prompt (transcript) and description prompt (style and speaker), instead of directly supporting human instruction. This division is less natural in form and does not align with other AIGC models. Secondly, the practice of utilizing an independent description prompt to model speech style, without considering the transcript content, restricts the ability to control speech at a fine-grained level. To address these limitations, we propose VoxInstruct, a novel unified multilingual codec language modeling framework that extends traditional text-to-speech tasks into a general human instruction-to-speech task. Our approach enhances the expressiveness of human instruction-guided speech generation and aligns the speech generation paradigm with other modalities. To enable the model to automatically extract the content of synthesized speech from raw text instructions, we introduce speech semantic tokens as an intermediate representation for instruction-to-content guidance. We also incorporate multiple Classifier-Free Guidance (CFG) strategies into our codec language model, which strengthens the generated speech following human instructions. Furthermore, our model architecture and training strategies allow for the simultaneous support of combining speech prompt and descriptive human instruction for expressive speech synthesis, which is a first-of-its-kind attempt. Codes, models and demos are at: https://github.com/thuhcsi/VoxInstruct.

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.