To reduce a model size but retain performance, we often rely on knowledge distillation (KD) which transfers knowledge from a large "teacher" model to a smaller "student" model. However, KD on multimodal datasets such as vision-language tasks is relatively unexplored, and digesting multimodal information is challenging since different modalities present different types of information. In this paper, we perform a large-scale empirical study to investigate the importance and effects of each modality in knowledge distillation. Furthermore, we introduce a multimodal knowledge distillation framework, modality-specific distillation (MSD), to transfer knowledge from a teacher on multimodal tasks by learning the teacher's behavior within each modality. The idea aims at mimicking a teacher's modality-specific predictions by introducing auxiliary loss terms for each modality. Furthermore, because each modality has different saliency for predictions, we define saliency scores for each modality and investigate saliency-based weighting schemes for the auxiliary losses. We further study a weight learning approach to learn the optimal weights on these loss terms. In our empirical analysis, we examine the saliency of each modality in KD, demonstrate the effectiveness of the weighting scheme in MSD, and show that it achieves better performance than KD on four multimodal datasets.
Huazhong University of Science and Technology (HUST)（华中科技大学）；Nanjing University of Science and Technology（南京理工大学）；Nanyang Technological University（南洋理工大学）；National University of Singapore（新加坡国立大学）；Institute of North Electronic Equipment（北方电子设备研究所）
The key challenge of zero-shot learning (ZSL) is how to infer the latent semantic knowledge between visual and attribute features on seen classes, and thus achieving a desirable knowledge transfer to unseen classes. Prior works either simply align the global features of an image with its associated class semantic vector or utilize unidirectional attention to learn the limited latent semantic representations, which could not effectively discover the intrinsic semantic knowledge e.g., attribute semantics) between visual and attribute features. To solve the above dilemma, we propose a Mutually Se…
As the development of neural networks, more and more deep neural networks are adopted in various tasks, such as image classification. However, as the huge computational overhead, these networks could not be applied on mobile devices or other low latency scenes. To address this dilemma, multi-classifier convolutional network is proposed to allow faster inference via early classifiers with the corresponding classifiers. These networks utilize sophisticated designing to increase the early classifier accuracy. However, naively training the multi-classifier network could hurt the performance (accur…
Temporal Saliency-Guided Distillation: A Scalable Framework for Distilling Video Datasets
Harbin Institute of Technology, Shenzhen（哈尔滨工业大学（深圳））；Peng Cheng Laboratory（鹏城实验室）；Tsinghua Shenzhen International Graduate School（清华大学深圳国际研究生院）；Shanghai Jiao Tong University（上海交通大学）
Insight SFI Research Centre for Data Analytics（Insight SFI数据分析研究中心）；Dublin City University（都柏林城市大学）；University of Catania（卡塔尼亚大学）
Video saliency prediction has recently attracted attention of the research community, as it is an upstream task for several practical applications. However, current solutions are particularly computationally demanding, especially due to the wide usage of spatio-temporal 3D convolutions. We observe that, while different model architectures achieve similar performance on benchmarks, visual variations between predicted saliency maps are still significant. Inspired by this intuition, we propose a lightweight model that employs multiple simple heterogeneous decoders and adopts several practical app…
Beihang University（北京航空航天大学）；Peng Cheng Laboratory（鹏城实验室）；Institute of Information Engineering, Chinese Academy of Sciences（中国科学院信息工程研究所）；University of Chinese Academy of Sciences（中国科学院大学）
The performance of video saliency estimation techniques has achieved significant advances along with the rapid development of Convolutional Neural Networks (CNNs). However, devices like cameras and drones may have limited computational capability and storage space so that the direct deployment of complex deep saliency models becomes infeasible. To address this problem, this paper proposes a dynamic saliency estimation approach for aerial videos via spatiotemporal knowledge distillation. In this approach, five components are involved, including two teachers, two students and the desired spatiot…
Mutually Causal Semantic Distillation Network for Zero-Shot Learning
Huazhong University of Science and Technology（华中科技大学）；National Anti-Counterfeit Engineering Research Center（国家防伪工程技术研究中心）；Nanjing University of Science and Technology（南京理工大学）
School of Computer Science, Wuhan University（武汉大学计算机学院）；School of Computer Science, The University of Sydney（悉尼大学计算机学院）；School of Computer Science and Engineering, Nanyang Technological University（南洋理工大学计算机科学与工程学院）
Robust Saliency-Aware Distillation for Few-shot Fine-grained Visual Recognition
School of Computer Science and Engineering, South China University of Technology（华南理工大学计算机科学与工程学院）；Brain and Affective Cognitive Research Center, Pazhou Lab（琶洲实验室脑与情感认知研究中心）；Engineering Research Center of the Ministry of Education on Health Intelligent Perception and Paralleled Digital-Human（教育部健康智能感知与平行数字人工程研究中心）
MSD-KMamba: Bidirectional Spatial-Aware Multi-Modal 3D Brain Segmentation via Multi-scale Self-Distilled Fusion Strategy
Key Laboratory of Intelligent Computing and Signal Processing, Ministry of Education, Anhui University（智能计算与信号处理重点实验室，教育部，安徽大学）；Key Laboratory of Smart Manufacturing in Energy Chemical Process, Ministry of Education, East China University of Science and Technology（能源化工过程智能制造重点实验室，教育部，东华大学）
We design a real-time portrait matting pipeline for everyday use, particularly for "virtual backgrounds" in video conferences. Existing segmentation and matting methods prioritize accuracy and quality over throughput and efficiency, and our pipeline enables trading off a controllable amount of accuracy for better throughput by leveraging online distillation on the input video stream. We construct our own dataset of simulated video calls in various scenarios, and show that our approach delivers a 5x speedup over a saliency detection based pipeline in a non-GPU accelerated setting while deliveri…
Light field data exhibit favorable characteristics conducive to saliency detection. The success of learning-based light field saliency detection is heavily dependent on how a comprehensive dataset can be constructed for higher generalizability of models, how high dimensional light field data can be effectively exploited, and how a flexible model can be designed to achieve versatility for desktop computers and mobile devices. To answer these questions, first we introduce a large-scale dataset to enable versatile applications for RGB, RGB-D and light field saliency detection, containing 102 clas…
UMSPU: Universal Multi-Size Phase Unwrapping via Mutual Self-Distillation and Adaptive Boosting Ensemble Segmenters
the School of Electronic Information and Electrical Engineering（电子信息与电气工程学院）；the Institute of Marine Equipment（海洋装备研究院）；the Institute of Medical Robotics（医学机器人研究院）
In this paper, we propose an approach for filter-level pruning with hierarchical knowledge distillation based on the teacher, teaching-assistant, and student framework. Our method makes use of teaching assistants at intermediate pruning levels that share the same architecture and weights as the target student. We propose to prune each model independently using the gradient information from its corresponding teacher. By considering the relative sizes of each student-teacher pair, this formulation provides a natural trade-off between the capacity gap for knowledge distillation and the bias of th…
In video understanding, most cross-modal knowledge distillation (KD) methods are tailored for classification tasks, focusing on the discriminative representation of the trimmed videos. However, action detection requires not only categorizing actions, but also localizing them in untrimmed videos. Therefore, transferring knowledge pertaining to temporal relations is critical for this task which is missing in the previous cross-modal KD frameworks. To this end, we aim at learning an augmented RGB representation for action detection, taking advantage of additional modalities at training time throu…
Department of Information Engineering Hong Kong；Department of Information Engineering
Since the wide employment of deep learning frameworks in video salient object detection, the accuracy of the recent approaches has made stunning progress. These approaches mainly adopt the sequential modules, based on optical flow or recurrent neural network (RNN), to learn robust spatiotemporal features. These modules are effective but significantly increase the computational burden of the corresponding deep models. In this paper, to simplify the network and maintain the accuracy, we present a lightweight network tailored for video salient object detection through the spatiotemporal knowledge…
Multi-Dataset Cross-Domain Knowledge Distillation for Unified Medical Image Segmentation, Classification, and Detection
Faculty of Mathematics and Computer Science, University of Bucharest（布加勒斯特大学数学与计算机科学学院）；Gheorghe Mihoc-Caius Iacob Institute of Mathematical Statistics and Applied Mathematics of the Romanian Academy（罗马尼亚科学院格奥尔基·米霍茨-凯奥斯·伊阿科夫学院数学统计与应用数学研究所）
Shanghai Em-Data Technology Co.（上海翼眸科技有限公司）；The University of Adelaide（阿德莱德大学）；Monash University（莫纳什大学）；Baidu Inc.（百度公司）
Knowledge distillation (KD) has been proven to be a simple and effective tool for training compact models. Almost all KD variants for dense prediction tasks align the student and teacher networks' feature maps in the spatial domain, typically by minimizing point-wise and/or pair-wise discrepancy. Observing that in semantic segmentation, some layers' feature activations of each channel tend to encode saliency of scene categories (analogue to class activation mapping), we propose to align features channel-wise between the student and teacher networks. To this end, we first transform the feature …
SenseTime X-Lab（商汤科技X-Lab）；CUHK - SenseTime Joint Lab, The Chinese University of Hong Kong（香港中文大学-商汤科技联合实验室）
Learning group representation is a commonly concerned issue in tasks where the basic unit is a group, set, or sequence. Previously, the research community tries to tackle it by aggregating the elements in a group based on an indicator either defined by humans such as the quality and saliency, or generated by a black box such as the attention score. This article provides a more essential and explicable view. We claim the most significant indicator to show whether the group representation can be benefited from one of its element is not the quality or an inexplicable score, but the discriminabili…
Predicting human fixations from images has recently seen large improvements by leveraging deep representations which were pretrained for object recognition. However, as we show in this paper, these networks are highly overparameterized for the task of fixation prediction. We first present a simple yet principled greedy pruning method which we call Fisher pruning. Through a combination of knowledge distillation and Fisher pruning, we obtain much more runtime-efficient architectures for saliency prediction, achieving a 10x speedup for the same AUC performance as a state of the art network on the…
Cross-Modal Knowledge Distillation for PET-Free Amyloid-Beta Detection from MRI
Dublin City University（都柏林城市大学）；Insight Research Ireland Centre for Data Analytics（爱尔兰Insight研究数据分析师中心）；St. Vincent’s University Hospital（圣文森医院）；University College Dublin（都柏林大学）
Boosting Medical Vision-Language Pretraining via Momentum Self-Distillation under Limited Computing Resources
Faculty of Information Technology, University of Science, Ho Chi Minh City, Vietnam（信息科技学院，科学大学，胡志明市，越南）；Vietnam National University, Ho Chi Minh City, Vietnam（越南国家大学，胡志明市，越南）
Most existing CNN-based salient object detection methods can identify local segmentation details like hair and animal fur, but often misinterpret the real saliency due to the lack of global contextual information caused by the subjectiveness of the SOD task and the locality of convolution layers. Moreover, due to the unrealistically expensive labeling costs, the current existing SOD datasets are insufficient to cover the real data distribution. The limitation and bias of the training data add additional difficulty to fully exploring the semantic association between object-to-object and object-…
