Li Su

Yunbin Tu, Liang Li, Li Su et al.

Change captioning aims to describe the difference between a pair of similar images. Its key challenge is how to learn a stable difference representation under pseudo changes caused by viewpoint change. In this paper, we address this by proposing a self-supervised cross-view representation reconstruction (SCORER) network. Concretely, we first design a multi-head token-wise matching to model relationships between cross-view features from similar/dissimilar images. Then, by maximizing cross-view contrastive alignment of two similar images, SCORER learns two view-invariant image representations in a self-supervised way. Based on these, we reconstruct the representations of unchanged objects by cross-attention, thus learning a stable difference representation for caption generation. Further, we devise a cross-modal backward reasoning to improve the quality of caption. This module reversely models a ``hallucination'' representation with the caption and ``before'' representation. By pushing it closer to the ``after'' representation, we enforce the caption to be informative about the difference in a self-supervised manner. Extensive experiments show our method achieves the state-of-the-art results on four datasets. The code is available at https://github.com/tuyunbin/SCORER.

Yunbin Tu, Liang Li, Li Su et al.

Change captioning is to describe the semantic change between a pair of similar images in natural language. It is more challenging than general image captioning, because it requires capturing fine-grained change information while being immune to irrelevant viewpoint changes, and solving syntax ambiguity in change descriptions. In this paper, we propose a neighborhood contrastive transformer to improve the model's perceiving ability for various changes under different scenes and cognition ability for complex syntax structure. Concretely, we first design a neighboring feature aggregating to integrate neighboring context into each feature, which helps quickly locate the inconspicuous changes under the guidance of conspicuous referents. Then, we devise a common feature distilling to compare two images at neighborhood level and extract common properties from each image, so as to learn effective contrastive information between them. Finally, we introduce the explicit dependencies between words to calibrate the transformer decoder, which helps better understand complex syntax structure during training. Extensive experimental results demonstrate that the proposed method achieves the state-of-the-art performance on three public datasets with different change scenarios. The code is available at https://github.com/tuyunbin/NCT.

Yunbin Tu, Liang Li, Li Su et al.

Change captioning aims to succinctly describe the semantic change between a pair of similar images, while being immune to distractors (illumination and viewpoint changes). Under these distractors, unchanged objects often appear pseudo changes about location and scale, and certain objects might overlap others, resulting in perturbational and discrimination-degraded features between two images. However, most existing methods directly capture the difference between them, which risk obtaining error-prone difference features. In this paper, we propose a distractors-immune representation learning network that correlates the corresponding channels of two image representations and decorrelates different ones in a self-supervised manner, thus attaining a pair of stable image representations under distractors. Then, the model can better interact them to capture the reliable difference features for caption generation. To yield words based on the most related difference features, we further design a cross-modal contrastive regularization, which regularizes the cross-modal alignment by maximizing the contrastive alignment between the attended difference features and generated words. Extensive experiments show that our method outperforms the state-of-the-art methods on four public datasets. The code is available at https://github.com/tuyunbin/DIRL.

Sangeon Yong, Li Su, Juhan Nam

Note-level automatic music transcription is one of the most representative music information retrieval (MIR) tasks and has been studied for various instruments to understand music. However, due to the lack of high-quality labeled data, transcription of many instruments is still a challenging task. In particular, in the case of singing, it is difficult to find accurate notes due to its expressiveness in pitch, timbre, and dynamics. In this paper, we propose a method of finding note onsets of singing voice more accurately by leveraging the linguistic characteristics of singing, which are not seen in other instruments. The proposed model uses mel-scaled spectrogram and phonetic posteriorgram (PPG), a frame-wise likelihood of phoneme, as an input of the onset detection network while PPG is generated by the pre-trained network with singing and speech data. To verify how linguistic features affect onset detection, we compare the evaluation results through the dataset with different languages and divide onset types for detailed analysis. Our approach substantially improves the performance of singing transcription and therefore emphasizes the importance of linguistic features in singing analysis.

Haochuan Cui, Junjie Sheng, Bo Jin et al.

With the rapid development of cloud computing, virtual machine scheduling has become one of the most important but challenging issues for the cloud computing community, especially for practical heterogeneous request sequences. By analyzing the impact of request heterogeneity on some popular heuristic schedulers, it can be found that existing scheduling algorithms can not handle the request heterogeneity properly and efficiently. In this paper, a plug-and-play virtual machine scheduling intensifier, called Resource Assigner (ReAssigner), is proposed to enhance the scheduling efficiency of any given scheduler for heterogeneous requests. The key idea of ReAssigner is to pre-assign roles to physical resources and let resources of the same role form a virtual cluster to handle homogeneous requests. ReAssigner can cooperate with arbitrary schedulers by restricting their scheduling space to virtual clusters. With evaluations on the real dataset from Huawei Cloud, the proposed ReAssigner achieves significant scheduling performance improvement compared with some state-of-the-art scheduling methods.

Hongcheng Gao, Jiashu Qu, Jingyi Tang et al. · tsinghua

The hallucination of large multimodal models (LMMs), providing responses that appear correct but are actually incorrect, limits their reliability and applicability. This paper aims to study the hallucination problem of LMMs in video modality, which is dynamic and more challenging compared to static modalities like images and text. From this motivation, we first present a comprehensive benchmark termed HAVEN for evaluating hallucinations of LMMs in video understanding tasks. It is built upon three dimensions, i.e., hallucination causes, hallucination aspects, and question formats, resulting in 6K questions. Then, we quantitatively study 7 influential factors on hallucinations, e.g., duration time of videos, model sizes, and model reasoning, via experiments of 16 LMMs on the presented benchmark. In addition, inspired by recent thinking models like OpenAI o1, we propose a video-thinking model to mitigate the hallucinations of LMMs via supervised reasoning fine-tuning (SRFT) and direct preference optimization (TDPO)-- where SRFT enhances reasoning capabilities while TDPO reduces hallucinations in the thinking process. Extensive experiments and analyses demonstrate the effectiveness. Remarkably, it improves the baseline by 7.65% in accuracy on hallucination evaluation and reduces the bias score by 4.5%. The code and data are public at https://github.com/Hongcheng-Gao/HAVEN.

Minheng Chen, Jing Zhang, Tong Chen et al.

Alzheimer's disease (AD) and Lewy body dementia (LBD) present overlapping clinical features yet require distinct diagnostic strategies. While neuroimaging-based brain network analysis is promising, atlas-based representations may obscure individualized anatomy. Gyral folding-based networks using three-hinge gyri provide a biologically grounded alternative, but inter-individual variability in cortical folding results in inconsistent landmark correspondence and highly irregular network sizes, violating the fixed-topology and node-alignment assumptions of most existing graph learning methods, particularly in clinical datasets where pathological changes further amplify anatomical heterogeneity. We therefore propose a probability-invariant random-walk-based framework that classifies individualized gyral folding networks without explicit node alignment. Cortical similarity networks are built from local morphometric features and represented by distributions of anonymized random walks, with an anatomy-aware encoding that preserves permutation invariance. Experiments on a large clinical cohort of AD and LBD subjects show consistent improvements over existing gyral folding and atlas-based models, demonstrating robustness and potential for dementia diagnosis.

Xiong Xiong, Li Su, Jinguo Huang et al.

Objective: Motor Imagery (MI) serves as a crucial experimental paradigm within the realm of Brain Computer Interfaces (BCIs), aiming to decoding motor intentions from electroencephalogram (EEG) signals. Method: Drawing inspiration from Riemannian geometry and Cross-Frequency Coupling (CFC), this paper introduces a novel approach termed Riemann Tangent Space Mapping using Dichotomous Filter Bank with Convolutional Neural Network (DFBRTS) to enhance the representation quality and decoding capability pertaining to MI features. DFBRTS first initiates the process by meticulously filtering EEG signals through a Dichotomous Filter Bank, structured in the fashion of a complete binary tree. Subsequently, it employs Riemann Tangent Space Mapping to extract salient EEG signal features within each sub-band. Finally, a lightweight convolutional neural network is employed for further feature extraction and classification, operating under the joint supervision of cross-entropy and center loss. To validate the efficacy, extensive experiments were conducted using DFBRTS on two well-established benchmark datasets: the BCI competition IV 2a (BCIC-IV-2a) dataset and the OpenBMI dataset. The performance of DFBRTS was benchmarked against several state-of-the-art MI decoding methods, alongside other Riemannian geometry-based MI decoding approaches. Results: DFBRTS significantly outperforms other MI decoding algorithms on both datasets, achieving a remarkable classification accuracy of 78.16% for four-class and 71.58% for two-class hold-out classification, as compared to the existing benchmarks.

Jingjing Zhou, Gaoxiang Cong, Li Su et al.

Large Reasoning Models (LRMs) have advanced automated multi-step reasoning, but their ability to generate complex Chain-of-Thought (CoT) trajectories introduces severe privacy risks, as sensitive information may be deeply embedded throughout the reasoning process. Existing Large Language Models (LLMs) unlearning approaches that typically focus on modifying only final answers are insufficient for LRMs, as they fail to remove sensitive content from intermediate steps, leading to persistent privacy leakage and degraded security. To address these challenges, we propose Sensitive Trajectory Regulation (STaR), a parameter-free, inference-time unlearning framework that achieves robust privacy protection throughout the reasoning process. Specifically, we first identify sensitive content via semantic-aware detection. Then, we inject global safety constraints through secure prompt prefix. Next, we perform trajectory-aware suppression to dynamically block sensitive content across the entire reasoning chain. Finally, we apply token-level adaptive filtering to prevent both exact and paraphrased sensitive tokens during generation. Furthermore, to overcome the inadequacies of existing evaluation protocols, we introduce two metrics: Multi-Decoding Consistency Assessment (MCS), which measures the consistency of unlearning across diverse decoding strategies, and Multi-Granularity Membership Inference Attack (MIA) Evaluation, which quantifies privacy protection at both answer and reasoning-chain levels. Experiments on the R-TOFU benchmark demonstrate that STaR achieves comprehensive and stable unlearning with minimal utility loss, setting a new standard for privacy-preserving reasoning in LRMs.

Yu-Fen Huang, Nikki Moran, Simon Coleman et al.

In cross-modal music processing, translation between visual, auditory, and semantic content opens up new possibilities as well as challenges. The construction of such a transformative scheme depends upon a benchmark corpus with a comprehensive data infrastructure. In particular, the assembly of a large-scale cross-modal dataset presents major challenges. In this paper, we present the MOSA (Music mOtion with Semantic Annotation) dataset, which contains high quality 3-D motion capture data, aligned audio recordings, and note-by-note semantic annotations of pitch, beat, phrase, dynamic, articulation, and harmony for 742 professional music performances by 23 professional musicians, comprising more than 30 hours and 570 K notes of data. To our knowledge, this is the largest cross-modal music dataset with note-level annotations to date. To demonstrate the usage of the MOSA dataset, we present several innovative cross-modal music information retrieval (MIR) and musical content generation tasks, including the detection of beats, downbeats, phrase, and expressive contents from audio, video and motion data, and the generation of musicians' body motion from given music audio. The dataset and codes are available alongside this publication (https://github.com/yufenhuang/MOSA-Music-mOtion-and-Semantic-Annotation-dataset).

Zhaobo Qi, Shuhui Wang, Chi Su et al.

Event analysis in untrimmed videos has attracted increasing attention due to the application of cutting-edge techniques such as CNN. As a well studied property for CNN-based models, the receptive field is a measurement for measuring the spatial range covered by a single feature response, which is crucial in improving the image categorization accuracy. In video domain, video event semantics are actually described by complex interaction among different concepts, while their behaviors vary drastically from one video to another, leading to the difficulty in concept-based analytics for accurate event categorization. To model the concept behavior, we study temporal concept receptive field of concept-based event representation, which encodes the temporal occurrence pattern of different mid-level concepts. Accordingly, we introduce temporal dynamic convolution (TDC) to give stronger flexibility to concept-based event analytics. TDC can adjust the temporal concept receptive field size dynamically according to different inputs. Notably, a set of coefficients are learned to fuse the results of multiple convolutions with different kernel widths that provide various temporal concept receptive field sizes. Different coefficients can generate appropriate and accurate temporal concept receptive field size according to input videos and highlight crucial concepts. Based on TDC, we propose the temporal dynamic concept modeling network (TDCMN) to learn an accurate and complete concept representation for efficient untrimmed video analysis. Experiment results on FCVID and ActivityNet show that TDCMN demonstrates adaptive event recognition ability conditioned on different inputs, and improve the event recognition performance of Concept-based methods by a large margin. Code is available at https://github.com/qzhb/TDCMN.

Shaofei Cai, Liang Li, Jincan Deng et al.

Graph neural networks (GNNs) emerged recently as a standard toolkit for learning from data on graphs. Current GNN designing works depend on immense human expertise to explore different message-passing mechanisms, and require manual enumeration to determine the proper message-passing depth. Inspired by the strong searching capability of neural architecture search (NAS) in CNN, this paper proposes Graph Neural Architecture Search (GNAS) with novel-designed search space. The GNAS can automatically learn better architecture with the optimal depth of message passing on the graph. Specifically, we design Graph Neural Architecture Paradigm (GAP) with tree-topology computation procedure and two types of fine-grained atomic operations (feature filtering and neighbor aggregation) from message-passing mechanism to construct powerful graph network search space. Feature filtering performs adaptive feature selection, and neighbor aggregation captures structural information and calculates neighbors' statistics. Experiments show that our GNAS can search for better GNNs with multiple message-passing mechanisms and optimal message-passing depth. The searched network achieves remarkable improvement over state-of-the-art manual designed and search-based GNNs on five large-scale datasets at three classical graph tasks. Codes can be found at https://github.com/phython96/GNAS-MP.

Jing Zhang, Xiaowei Yu, Tong Chen et al.

The Lewy body dementia (LBD) is the second most common neurodegenerative dementia after Alzheimer's disease (AD). Early differentiation between AD and LBD is crucial because they require different treatment approaches, but this is challenging due to significant clinical overlap, heterogeneity, complex pathogenesis, and the rarity of LBD. While recent advances in artificial intelligence (AI) demonstrate powerful learning capabilities and offer new hope for accurate diagnosis, existing methods primarily focus on designing "neural-level networks". Our work represents a pioneering effort in modeling system-level artificial neural network called BrainNet-MoE for brain modeling and diagnosing. Inspired by the brain's hierarchical organization of bottom-up sensory integration and top-down control, we design a set of disease-specific expert groups to process brain sub-network under different condition, A disease gate mechanism guides the specializa-tion of expert groups, while a transformer layer enables communication be-tween all sub-networks, generating a comprehensive whole-brain represen-tation for downstream disease classification. Experimental results show superior classification accuracy with interpretable insights into how brain sub-networks contribute to different neurodegenerative conditions.

Xiaowei Yu, Jing Zhang, Tong Chen et al.

Lewy Body Disease (LBD) is a common yet understudied form of dementia that imposes a significant burden on public health. It shares clinical similarities with Alzheimer's disease (AD), as both progress through stages of normal cognition, mild cognitive impairment, and dementia. A major obstacle in LBD diagnosis is data scarcity, which limits the effectiveness of deep learning. In contrast, AD datasets are more abundant, offering potential for knowledge transfer. However, LBD and AD data are typically collected from different sites using different machines and protocols, resulting in a distinct domain shift. To effectively leverage AD data while mitigating domain shift, we propose a Transferability Aware Transformer (TAT) that adapts knowledge from AD to enhance LBD diagnosis. Our method utilizes structural connectivity (SC) derived from structural MRI as training data. Built on the attention mechanism, TAT adaptively assigns greater weights to disease-transferable features while suppressing domain-specific ones, thereby reducing domain shift and improving diagnostic accuracy with limited LBD data. The experimental results demonstrate the effectiveness of TAT. To the best of our knowledge, this is the first study to explore domain adaptation from AD to LBD under conditions of data scarcity and domain shift, providing a promising framework for domain-adaptive diagnosis of rare diseases.

Ting-Kang Wang, Yueh-Po Peng, Li Su et al.

While automatic music transcription is well-established in music information retrieval, most models are limited to transcribing pitch and timing information from audio, and thus omit crucial expressive and instrument-specific nuances. One example is playing technique on the violin, which affords its distinct palette of timbres for maximal emotional impact. Here, we propose VioPTT (Violin Playing Technique-aware Transcription), a lightweight, end-to-end model that directly transcribes violin playing technique in addition to pitch onset and offset. Furthermore, we release MOSA-VPT, a novel, high-quality synthetic violin playing technique dataset to circumvent the need for manually labeled annotations. Leveraging this dataset, our model demonstrated strong generalization to real-world note-level violin technique recordings in addition to achieving state-of-the-art transcription performance. To our knowledge, VioPTT is the first to jointly combine violin transcription and playing technique prediction within a unified framework.

Jing Zhang, Xiaowei Yu, Minheng Chen et al.

Integrating brain imaging data with clinical reports offers a valuable opportunity to leverage complementary multimodal information for more effective and timely diagnosis in practical clinical settings. This approach has gained significant attention in brain disorder research, yet a key challenge remains: how to effectively link objective imaging data with subjective text-based reports, such as doctors' notes. In this work, we propose a novel framework that aligns brain connectomes with clinical reports in a shared cross-modal latent space at both the subject and connectome levels, thereby enhancing representation learning. The key innovation of our approach is that we treat brain subnetworks as tokens of imaging data, rather than raw image patches, to align with word tokens in clinical reports. This enables a more efficient identification of system-level associations between neuroimaging findings and clinical observations, which is critical since brain disorders often manifest as network-level abnormalities rather than isolated regional alterations. We applied our method to mild cognitive impairment (MCI) using the ADNI dataset. Our approach not only achieves state-of-the-art predictive performance but also identifies clinically meaningful connectome-text pairs, offering new insights into the early mechanisms of Alzheimer's disease and supporting the development of clinically useful multimodal biomarkers.

Yueh-Po Peng, Vincent K. M. Cheung, Li Su

A fundamental challenge in neuroscience is to decode mental states from brain activity. While functional magnetic resonance imaging (fMRI) offers a non-invasive approach to capture brain-wide neural dynamics with high spatial precision, decoding from fMRI data -- particularly from task-evoked activity -- remains challenging due to its high dimensionality, low signal-to-noise ratio, and limited within-subject data. Here, we leverage recent advances in computer vision and propose STDA-SwiFT, a transformer-based model that learns transferable representations from large-scale fMRI datasets via spatial-temporal divided attention and self-supervised contrastive learning. Using pretrained voxel-wise representations from 995 subjects in the Human Connectome Project (HCP), we show that our model substantially improves downstream decoding performance of task-evoked activity across multiple sensory and cognitive domains, even with minimal data preprocessing. We demonstrate performance gains from larger receptor fields afforded by our memory-efficient attention mechanism, as well as the impact of functional relevance in pretraining data when fine-tuning on small samples. Our work showcases transfer learning as a viable approach to harness large-scale datasets to overcome challenges in decoding brain activity from fMRI data.

Jun-You Wang, Li Su

We propose a pre-trained BERT-like model for symbolic music understanding that achieves competitive performance across a wide range of downstream tasks. To achieve this target, we design two novel pre-training objectives, namely token correction and pianoroll prediction. First, we sample a portion of note tokens and corrupt them with a limited amount of noise, and then train the model to denoise the corrupted tokens; second, we also train the model to predict bar-level and local pianoroll-derived representations from the corrupted note tokens. We argue that these objectives guide the model to better learn specific musical knowledge such as pitch intervals. For evaluation, we propose a benchmark that incorporates 12 downstream tasks ranging from chord estimation to symbolic genre classification. Results confirm the effectiveness of the proposed pre-training objectives on downstream tasks.

Yunbin Tu, Liang Li, Li Su et al.

Video has emerged as a favored multimedia format on the internet. To better gain video contents, a new topic HIREST is presented, including video retrieval, moment retrieval, moment segmentation, and step-captioning. The pioneering work chooses the pre-trained CLIP-based model for video retrieval, and leverages it as a feature extractor for other three challenging tasks solved in a multi-task learning paradigm. Nevertheless, this work struggles to learn the comprehensive cognition of user-preferred content, due to disregarding the hierarchies and association relations across modalities. In this paper, guided by the shallow-to-deep principle, we propose a query-centric audio-visual cognition (QUAG) network to construct a reliable multi-modal representation for moment retrieval, segmentation and step-captioning. Specifically, we first design the modality-synergistic perception to obtain rich audio-visual content, by modeling global contrastive alignment and local fine-grained interaction between visual and audio modalities. Then, we devise the query-centric cognition that uses the deep-level query to perform the temporal-channel filtration on the shallow-level audio-visual representation. This can cognize user-preferred content and thus attain a query-centric audio-visual representation for three tasks. Extensive experiments show QUAG achieves the SOTA results on HIREST. Further, we test QUAG on the query-based video summarization task and verify its good generalization.

Zhaobo Qi, Shuhui Wang, Chi Su et al.

Future activity anticipation is a challenging problem in egocentric vision. As a standard future activity anticipation paradigm, recursive sequence prediction suffers from the accumulation of errors. To address this problem, we propose a simple and effective Self-Regulated Learning framework, which aims to regulate the intermediate representation consecutively to produce representation that (a) emphasizes the novel information in the frame of the current time-stamp in contrast to previously observed content, and (b) reflects its correlation with previously observed frames. The former is achieved by minimizing a contrastive loss, and the latter can be achieved by a dynamic reweighing mechanism to attend to informative frames in the observed content with a similarity comparison between feature of the current frame and observed frames. The learned final video representation can be further enhanced by multi-task learning which performs joint feature learning on the target activity labels and the automatically detected action and object class tokens. SRL sharply outperforms existing state-of-the-art in most cases on two egocentric video datasets and two third-person video datasets. Its effectiveness is also verified by the experimental fact that the action and object concepts that support the activity semantics can be accurately identified.

Wei-Tsung Lu, Meng-Hsuan Wu, Yuh-Ming Chiu et al.

The subjective evaluation of music generation techniques has been mostly done with questionnaire-based listening tests while ignoring the perspectives from music composition, arrangement, and soundtrack editing. In this paper, we propose an editing test to evaluate users' editing experience of music generation models in a systematic way. To do this, we design a new music style transfer model combining the non-chronological inference architecture, autoregressive models and the Transformer, which serves as an improvement from the baseline model on the same style transfer task. Then, we compare the performance of the two models with a conventional listening test and the proposed editing test, in which the quality of generated samples is assessed by the amount of effort (e.g., the number of required keyboard and mouse actions) spent by users to polish a music clip. Results on two target styles indicate that the improvement over the baseline model can be reflected by the editing test quantitatively. Also, the editing test provides profound insights which are not accessible from usual listening tests. The major contribution of this paper is the systematic presentation of the editing test and the corresponding insights, while the proposed music style transfer model based on state-of-the-art neural networks represents another contribution.

Kin Wai Cheuk, Dorien Herremans, Li Su

Most of the current supervised automatic music transcription (AMT) models lack the ability to generalize. This means that they have trouble transcribing real-world music recordings from diverse musical genres that are not presented in the labelled training data. In this paper, we propose a semi-supervised framework, ReconVAT, which solves this issue by leveraging the huge amount of available unlabelled music recordings. The proposed ReconVAT uses reconstruction loss and virtual adversarial training. When combined with existing U-net models for AMT, ReconVAT achieves competitive results on common benchmark datasets such as MAPS and MusicNet. For example, in the few-shot setting for the string part version of MusicNet, ReconVAT achieves F1-scores of 61.0% and 41.6% for the note-wise and note-with-offset-wise metrics respectively, which translates into an improvement of 22.2% and 62.5% compared to the supervised baseline model. Our proposed framework also demonstrates the potential of continual learning on new data, which could be useful in real-world applications whereby new data is constantly available.

Yu-Te Wu, Yin-Jyun Luo, Tsung-Ping Chen et al.

We present and release Omnizart, a new Python library that provides a streamlined solution to automatic music transcription (AMT). Omnizart encompasses modules that construct the life-cycle of deep learning-based AMT, and is designed for ease of use with a compact command-line interface. To the best of our knowledge, Omnizart is the first transcription toolkit which offers models covering a wide class of instruments ranging from solo, instrument ensembles, percussion instruments to vocal, as well as models for chord recognition and beat/downbeat tracking, two music information retrieval (MIR) tasks highly related to AMT.

Jinjian Li, Chuandong Guo, Li Su et al.

Event cameras are novel neuromorphic vision sensors with ultrahigh temporal resolution and low latency, both in the order of microseconds. Instead of image frames, event cameras generate an asynchronous event stream of per-pixel intensity changes with precise timestamps. The resulting sparse data structure impedes applying many conventional computer vision techniques to event streams, and specific algorithms should be designed to leverage the information provided by event cameras. We propose a corner detection algorithm, eSUSAN, inspired by the conventional SUSAN (smallest univalue segment assimilating nucleus) algorithm for corner detection. The proposed eSUSAN extracts the univalue segment assimilating nucleus from the circle kernel based on the similarity across timestamps and distinguishes corner events by the number of pixels in the nucleus area. Moreover, eSUSAN is fast enough to be applied to CeleX-V, the event camera with the highest resolution available. Based on eSUSAN, we also propose the SE-Harris corner detector, which uses adaptive normalization based on exponential decay to quickly construct a local surface of active events and the event-based Harris detector to refine the corners identified by eSUSAN. We evaluated the proposed algorithms on a public dataset and CeleX-V data. Both eSUSAN and SE-Harris exhibit higher real-time performance than existing algorithms while maintaining high accuracy and tracking performance.

Yin-Jyun Luo, Yuen-Jen Lin, Li Su

Singing voice correction (SVC) is an appealing application for amateur singers. Commercial products automate SVC by snapping pitch contours to equal-tempered scales, which could lead to deadpan modifications. Together with the neglect of rhythmic errors, extensive manual corrections are still necessary. In this paper, we present a streamlined system to automate expressive SVC for both pitch and rhythmic errors. Particularly, we extend a previous work by integrating advanced techniques for singing voice separation (SVS) and vocal melody extraction. SVC is achieved by temporally aligning the source-target pair, followed by replacing pitch and rhythm of the source with those of the target. We evaluate the framework by a comparative study for melody extraction which involves both subjective and objective evaluations, whereby we investigate perceptual validity of the standard metrics through the lens of SVC. The results suggest that the high pitch accuracy obtained by the metrics does not signify good perceptual scores.

Cheng-Che Lee, Wan-Yi Lin, Yen-Ting Shih et al.

Music-to-visual style transfer is a challenging yet important cross-modal learning problem in the practice of creativity. Its major difference from the traditional image style transfer problem is that the style information is provided by music rather than images. Assuming that musical features can be properly mapped to visual contents through semantic links between the two domains, we solve the music-to-visual style transfer problem in two steps: music visualization and style transfer. The music visualization network utilizes an encoder-generator architecture with a conditional generative adversarial network to generate image-based music representations from music data. This network is integrated with an image style transfer method to accomplish the style transfer process. Experiments are conducted on WikiArt-IMSLP, a newly compiled dataset including Western music recordings and paintings listed by decades. By utilizing such a label to learn the semantic connection between paintings and music, we demonstrate that the proposed framework can generate diverse image style representations from a music piece, and these representations can unveil certain art forms of the same era. Subjective testing results also emphasize the role of the era label in improving the perceptual quality on the compatibility between music and visual content.

Hsuan-Kai Kao, Li Su

This paper presents a neural network model to generate virtual violinist's 3-D skeleton movements from music audio. Improved from the conventional recurrent neural network models for generating 2-D skeleton data in previous works, the proposed model incorporates an encoder-decoder architecture, as well as the self-attention mechanism to model the complicated dynamics in body movement sequences. To facilitate the optimization of self-attention model, beat tracking is applied to determine effective sizes and boundaries of the training examples. The decoder is accompanied with a refining network and a bowing attack inference mechanism to emphasize the right-hand behavior and bowing attack timing. Both objective and subjective evaluations reveal that the proposed model outperforms the state-of-the-art methods. To the best of our knowledge, this work represents the first attempt to generate 3-D violinists' body movements considering key features in musical body movement.

Yuen-Jen Lin, Hsuan-Kai Kao, Yih-Chih Tseng et al.

Virtual musicians have become a remarkable phenomenon in the contemporary multimedia arts. However, most of the virtual musicians nowadays have not been endowed with abilities to create their own behaviors, or to perform music with human musicians. In this paper, we firstly create a virtual violinist, who can collaborate with a human pianist to perform chamber music automatically without any intervention. The system incorporates the techniques from various fields, including real-time music tracking, pose estimation, and body movement generation. In our system, the virtual musician's behavior is generated based on the given music audio alone, and such a system results in a low-cost, efficient and scalable way to produce human and virtual musicians' co-performance. The proposed system has been validated in public concerts. Objective quality assessment approaches and possible ways to systematically improve the system are also discussed.

Sangeun Kum, Jing-Hua Lin, Li Su et al.

The lack of labeled data is a major obstacle in many music information retrieval tasks such as melody extraction, where labeling is extremely laborious or costly. Semi-supervised learning (SSL) provides a solution to alleviate the issue by leveraging a large amount of unlabeled data. In this paper, we propose an SSL method using teacher-student models for vocal melody extraction. The teacher model is pre-trained with labeled data and guides the student model to make identical predictions given unlabeled input in a self-training setting. We examine three setups of teacher-student models with different data augmentation schemes and loss functions. Also, considering the scarcity of labeled data in the test phase, we artificially generate large-scale testing data with pitch labels from unlabeled data using an analysis-synthesis method. The results show that the SSL method significantly increases the performance against supervised learning only and the improvement depends on the teacher-student models, the size of unlabeled data, the number of self-training iterations, and other training details. We also find that it is essential to ensure that the unlabeled audio has vocal parts. Finally, we show that the proposed SSL method enables a baseline convolutional recurrent neural network model to achieve performance comparable to state-of-the-arts.

Xuejing Liu, Liang Li, Shuhui Wang et al.

Weakly supervised referring expression grounding (REG) aims at localizing the referential entity in an image according to linguistic query, where the mapping between the image region (proposal) and the query is unknown in the training stage. In referring expressions, people usually describe a target entity in terms of its relationship with other contextual entities as well as visual attributes. However, previous weakly supervised REG methods rarely pay attention to the relationship between the entities. In this paper, we propose a knowledge-guided pairwise reconstruction network (KPRN), which models the relationship between the target entity (subject) and contextual entity (object) as well as grounds these two entities. Specifically, we first design a knowledge extraction module to guide the proposal selection of subject and object. The prior knowledge is obtained in a specific form of semantic similarities between each proposal and the subject/object. Second, guided by such knowledge, we design the subject and object attention module to construct the subject-object proposal pairs. The subject attention excludes the unrelated proposals from the candidate proposals. The object attention selects the most suitable proposal as the contextual proposal. Third, we introduce a pairwise attention and an adaptive weighting scheme to learn the correspondence between these proposal pairs and the query. Finally, a pairwise reconstruction module is used to measure the grounding for weakly supervised learning. Extensive experiments on four large-scale datasets show our method outperforms existing state-of-the-art methods by a large margin.

Zhe Wu, Li Su, Qingming Huang

Existing state-of-the-art salient object detection networks rely on aggregating multi-level features of pre-trained convolutional neural networks (CNNs). Compared to high-level features, low-level features contribute less to performance but cost more computations because of their larger spatial resolutions. In this paper, we propose a novel Cascaded Partial Decoder (CPD) framework for fast and accurate salient object detection. On the one hand, the framework constructs partial decoder which discards larger resolution features of shallower layers for acceleration. On the other hand, we observe that integrating features of deeper layers obtain relatively precise saliency map. Therefore we directly utilize generated saliency map to refine the features of backbone network. This strategy efficiently suppresses distractors in the features and significantly improves their representation ability. Experiments conducted on five benchmark datasets exhibit that the proposed model not only achieves state-of-the-art performance but also runs much faster than existing models. Besides, the proposed framework is further applied to improve existing multi-level feature aggregation models and significantly improve their efficiency and accuracy.

Chin-Yun Yu, Li Su

We propose the multi-layered cepstrum (MLC) method to estimate multiple fundamental frequencies (MF0) of a signal under challenging contamination such as high-pass filter noise. Taking the operation of cepstrum (i.e., Fourier transform, filtering, and nonlinear activation) recursively, MLC is shown as an efficient method to enhance MF0 saliency in a step-by-step manner. Evaluation on a real-world polyphonic music dataset under both normal and low-fidelity conditions demonstrates the potential of MLC.

Chien-Yu Lu, Min-Xin Xue, Chia-Che Chang et al.

Style transfer of polyphonic music recordings is a challenging task when considering the modeling of diverse, imaginative, and reasonable music pieces in the style different from their original one. To achieve this, learning stable multi-modal representations for both domain-variant (i.e., style) and domain-invariant (i.e., content) information of music in an unsupervised manner is critical. In this paper, we propose an unsupervised music style transfer method without the need for parallel data. Besides, to characterize the multi-modal distribution of music pieces, we employ the Multi-modal Unsupervised Image-to-Image Translation (MUNIT) framework in the proposed system. This allows one to generate diverse outputs from the learned latent distributions representing contents and styles. Moreover, to better capture the granularity of sound, such as the perceptual dimensions of timbre and the nuance in instrument-specific performance, cognitively plausible features including mel-frequency cepstral coefficients (MFCC), spectral difference, and spectral envelope, are combined with the widely-used mel-spectrogram into a timber-enhanced multi-channel input representation. The Relativistic average Generative Adversarial Networks (RaGAN) is also utilized to achieve fast convergence and high stability. We conduct experiments on bilateral style transfer tasks among three different genres, namely piano solo, guitar solo, and string quartet. Results demonstrate the advantages of the proposed method in music style transfer with improved sound quality and in allowing users to manipulate the output.

Tsung-Han Hsieh, Li Su, Yi-Hsuan Yang

Melody extraction in polyphonic musical audio is important for music signal processing. In this paper, we propose a novel streamlined encoder/decoder network that is designed for the task. We make two technical contributions. First, drawing inspiration from a state-of-the-art model for semantic pixel-wise segmentation, we pass through the pooling indices between pooling and un-pooling layers to localize the melody in frequency. We can achieve result close to the state-of-the-art with much fewer convolutional layers and simpler convolution modules. Second, we propose a way to use the bottleneck layer of the network to estimate the existence of a melody line for each time frame, and make it possible to use a simple argmax function instead of ad-hoc thresholding to get the final estimation of the melody line. Our experiments on both vocal melody extraction and general melody extraction validate the effectiveness of the proposed model.

Li Su

A patch-based convolutional neural network (CNN) model presented in this paper for vocal melody extraction in polyphonic music is inspired from object detection in image processing. The input of the model is a novel time-frequency representation which enhances the pitch contours and suppresses the harmonic components of a signal. This succinct data representation and the patch-based CNN model enable an efficient training process with limited labeled data. Experiments on various datasets show excellent speed and competitive accuracy comparing to other deep learning approaches.

Li Su

This paper presents a new approach in understanding how deep neural networks (DNNs) work by applying homomorphic signal processing techniques. Focusing on the task of multi-pitch estimation (MPE), this paper demonstrates the equivalence relation between a generalized cepstrum and a DNN in terms of their structures and functionality. Such an equivalence relation, together with pitch perception theories and the recently established rectified-correlations-on-a-sphere (RECOS) filter analysis, provide an alternative way in explaining the role of the nonlinear activation function and the multi-layer structure, both of which exist in a cepstrum and a DNN. To validate the efficacy of this new approach, a new feature designed in the same fashion is proposed for pitch salience function. The new feature outperforms the one-layer spectrum in the MPE task and, as predicted, it addresses the issue of the missing fundamental effect and also achieves better robustness to noise.

Li Su, Hau-tieng Wu

We study the real-time dynamics retrieval from a time series via the time-frequency (TF) analysis with the minimal latency guarantee. While different from the well-known intrinsic latency definition in the filter design, a rigorous definition of intrinsic latency for different time-frequency representations (TFR) is provided, including the short time Fourier transform (STFT), synchrosqeezing transform (SST) and reassignment method (RM). To achieve the minimal latency, a systematic method is proposed to construct an asymmetric window from a well-designed symmetric one based on the concept of minimum-phase, if the window satisfies some weak conditions. We theoretically show that the TFR determined by SST with the constructed asymmetric window does have a smaller intrinsic latency. Finally, the music onset detection problem is studied to show the strength of the proposed algorithm.

Chen-Yun Lin, Li Su, Hau-tieng Wu

We propose to combine cepstrum and nonlinear time-frequency (TF) analysis to study mutiple component oscillatory signals with time-varying frequency and amplitude and with time-varying non-sinusoidal oscillatory pattern. The concept of cepstrum is applied to eliminate the wave-shape function influence on the TF analysis, and we propose a new algorithm, named de-shape synchrosqueezing transform (de-shape SST). The mathematical model, adaptive non-harmonic model, is introduced and the de-shape SST algorithm is theoretically analyzed. In addition to simulated signals, several different physiological, musical and biological signals are analyzed to illustrate the proposed algorithm.