The application of machine learning to medical ultrasound videos of the heart, i.e., echocardiography, has recently gained traction with the availability of large public datasets. Traditional supervised tasks, such as ejection fraction regression, are now making way for approaches focusing more on the latent structure of data distributions, as well as generative methods. We propose a model trained exclusively by knowledge distillation, either on real or synthetical data, involving retrieving masks suggested by a teacher model. We achieve state-of-the-art (SOTA) values on the task of identifying end-diastolic and end-systolic frames. By training the model only on synthetic data, it reaches segmentation capabilities close to the performance when trained on real data with a significantly reduced number of weights. A comparison with the 5 main existing methods shows that our method outperforms the others in most cases. We also present a new evaluation method that does not require human annotation and instead relies on a large auxiliary model. We show that this method produces scores consistent with those obtained from human annotations. Relying on the integrated knowledge from a vast amount of records, this method overcomes certain inherent limitations of human annotator labeling. Code: https://github.com/GregoirePetit/EchoDFKD
Spatio-temporal prediction is a crucial research area in data-driven urban computing, with implications for transportation, public safety, and environmental monitoring. However, scalability and generalization challenges remain significant obstacles. Advanced models often rely on Graph Neural Networks to encode spatial and temporal correlations, but struggle with the increased complexity of large-scale datasets. The recursive GNN-based message passing schemes used in these models hinder their training and deployment in real-life urban sensing scenarios. Moreover, long-spanning large-scale spatio-temporal data introduce distribution shifts, necessitating improved generalization performance. To address these challenges, we propose a simple framework for spatio-temporal prediction - EasyST paradigm. It learns lightweight and robust Multi-Layer Perceptrons (MLPs) by effectively distilling knowledge from complex spatio-temporal GNNs. We ensure robust knowledge distillation by integrating the spatio-temporal information bottleneck with teacher-bounded regression loss, filtering out task-irrelevant noise and avoiding erroneous guidance. We further enhance the generalization ability of the student model by incorporating spatial and temporal prompts to provide downstream task contexts. Evaluation on three spatio-temporal datasets for urban computing tasks demonstrates that EasyST surpasses state-of-the-art approaches in terms of efficiency and accuracy. The implementation code is available at: https://github.com/HKUDS/EasyST.
Brain is one the most complex organs in the human body. Due to its complexity, classification of brain tumors still poses a significant challenge, making brain tumors a particularly serious medical issue. Techniques such as Machine Learning (ML) coupled with Magnetic Resonance Imaging (MRI) have paved the way for doctors and medical institutions to classify different types of tumors. However, these techniques suffer from limitations that violate patients privacy. Federated Learning (FL) has recently been introduced to solve such an issue, but the FL itself suffers from limitations like communication costs and dependencies on model architecture, forcing all models to have identical architectures. In this paper, we propose FedBrain-Distill, an approach that leverages Knowledge Distillation (KD) in an FL setting that maintains the users privacy and ensures the independence of FL clients in terms of model architecture. FedBrain-Distill uses an ensemble of teachers that distill their knowledge to a simple student model. The evaluation of FedBrain-Distill demonstrated high-accuracy results for both Independent and Identically Distributed (IID) and non-IID data with substantial low communication costs on the real-world Figshare brain tumor dataset. It is worth mentioning that we used Dirichlet distribution to partition the data into IID and non-IID data. All the implementation details are accessible through our Github repository.
Accurate 3D object detection is vital for automated driving. While lidar sensors are well suited for this task, they are expensive and have limitations in adverse weather conditions. 3+1D imaging radar sensors offer a cost-effective, robust alternative but face challenges due to their low resolution and high measurement noise. Existing 3+1D imaging radar datasets include radar and lidar data, enabling cross-modal model improvements. Although lidar should not be used during inference, it can aid the training of radar-only object detectors. We explore two strategies to transfer knowledge from the lidar to the radar domain and radar-only object detectors: 1. multi-stage training with sequential lidar point cloud thin-out, and 2. cross-modal knowledge distillation. In the multi-stage process, three thin-out methods are examined. Our results show significant performance gains of up to 4.2 percentage points in mean Average Precision with multi-stage training and up to 3.9 percentage points with knowledge distillation by initializing the student with the teacher's weights. The main benefit of these approaches is their applicability to other 3D object detection networks without altering their architecture, as we show by analyzing it on two different object detectors. Our code is available at https://github.com/rst-tu-dortmund/lerojd
Diffusion models (DMs) have demonstrated exceptional generative capabilities across various areas, while they are hindered by slow inference speeds and high computational demands during deployment. The most common way to accelerate DMs involves reducing the number of denoising steps during generation, achieved through faster sampling solvers or knowledge distillation (KD). In contrast to prior approaches, we propose a novel method that transfers the capability of large pretrained DMs to faster architectures. Specifically, we employ KD in a distinct manner to compress DMs by distilling their generative ability into more rapid variants. Furthermore, considering that the source data is either unaccessible or too enormous to store for current generative models, we introduce a new paradigm for their distillation without source data, termed Data-Free Knowledge Distillation for Diffusion Models (DKDM). Generally, our established DKDM framework comprises two main components: 1) a DKDM objective that uses synthetic denoising data produced by pretrained DMs to optimize faster DMs without source data, and 2) a dynamic iterative distillation method that flexibly organizes the synthesis of denoising data, preventing it from slowing down the optimization process as the generation is slow. To our knowledge, this is the first attempt at using KD to distill DMs into any architecture in a data-free manner. Importantly, our DKDM is orthogonal to most existing acceleration methods, such as denoising step reduction, quantization and pruning. Experiments show that our DKDM is capable of deriving 2x faster DMs with performance remaining on par with the baseline. Notably, our DKDM enables pretrained DMs to function as "datasets" for training new DMs.
This paper focuses on modern efficient training and inference technologies on foundation models and illustrates them from two perspectives: model and system design. Model and System Design optimize LLM training and inference from different aspects to save computational resources, making LLMs more efficient, affordable, and more accessible. The paper list repository is available at https://github.com/NoakLiu/Efficient-Foundation-Models-Survey.
With the rising demand for high-resolution (HR) images, No-Reference Image Quality Assessment (NR-IQA) gains more attention, as it can ecaluate image quality in real-time on mobile devices and enhance user experience. However, existing NR-IQA methods often resize or crop the HR images into small resolution, which leads to a loss of important details. And most of them are of high computational complexity, which hinders their application on mobile devices due to limited computational resources. To address these challenges, we propose MobileIQA, a novel approach that utilizes lightweight backbones to efficiently assess image quality while preserving image details through high-resolution input. MobileIQA employs the proposed multi-view attention learning (MAL) module to capture diverse opinions, simulating subjective opinions provided by different annotators during the dataset annotation process. The model uses a teacher model to guide the learning of a student model through knowledge distillation. This method significantly reduces computational complexity while maintaining high performance. Experiments demonstrate that MobileIQA outperforms novel IQA methods on evaluation metrics and computational efficiency. The code is available at https://github.com/chencn2020/MobileIQA.
Federated Class Continual Learning (FCCL) merges the challenges of distributed client learning with the need for seamless adaptation to new classes without forgetting old ones. The key challenge in FCCL is catastrophic forgetting, an issue that has been explored to some extent in Continual Learning (CL). However, due to privacy preservation requirements, some conventional methods, such as experience replay, are not directly applicable to FCCL. Existing FCCL methods mitigate forgetting by generating historical data through federated training of GANs or data-free knowledge distillation. However, these approaches often suffer from unstable training of generators or low-quality generated data, limiting their guidance for the model. To address this challenge, we propose a novel method of data replay based on diffusion models. Instead of training a diffusion model, we employ a pre-trained conditional diffusion model to reverse-engineer each class, searching the corresponding input conditions for each class within the model's input space, significantly reducing computational resources and time consumption while ensuring effective generation. Furthermore, we enhance the classifier's domain generalization ability on generated and real data through contrastive learning, indirectly improving the representational capability of generated data for real data. Comprehensive experiments demonstrate that our method significantly outperforms existing baselines. Code is available at https://github.com/jinglin-liang/DDDR.
Leveraging the capabilities of Knowledge Distillation (KD) strategies, we devise a strategy to fight the recent retraction of face recognition datasets. Given a pretrained Teacher model trained on a real dataset, we show that carefully utilising synthetic datasets, or a mix between real and synthetic datasets to distil knowledge from this teacher to smaller students can yield surprising results. In this sense, we trained 33 different models with and without KD, on different datasets, with different architectures and losses. And our findings are consistent, using KD leads to performance gains across all ethnicities and decreased bias. In addition, it helps to mitigate the performance gap between real and synthetic datasets. This approach addresses the limitations of synthetic data training, improving both the accuracy and fairness of face recognition models.
In this paper, we propose HiTSR, a hierarchical transformer model for reference-based image super-resolution, which enhances low-resolution input images by learning matching correspondences from high-resolution reference images. Diverging from existing multi-network, multi-stage approaches, we streamline the architecture and training pipeline by incorporating the double attention block from GAN literature. Processing two visual streams independently, we fuse self-attention and cross-attention blocks through a gating attention strategy. The model integrates a squeeze-and-excitation module to capture global context from the input images, facilitating long-range spatial interactions within window-based attention blocks. Long skip connections between shallow and deep layers further enhance information flow. Our model demonstrates superior performance across three datasets including SUN80, Urban100, and Manga109. Specifically, on the SUN80 dataset, our model achieves PSNR/SSIM values of 30.24/0.821. These results underscore the effectiveness of attention mechanisms in reference-based image super-resolution. The transformer-based model attains state-of-the-art results without the need for purpose-built subnetworks, knowledge distillation, or multi-stage training, emphasizing the potency of attention in meeting reference-based image super-resolution requirements.
As in school, one teacher to cover all subjects is insufficient to distill equally robust information to a student. Hence, each subject is taught by a highly specialised teacher. Following a similar philosophy, we propose a multiple specialized teacher framework to distill knowledge to a student network. In our approach, directed at face recognition use cases, we train four teachers on one specific ethnicity, leading to four highly specialized and biased teachers. Our strategy learns a project of these four teachers into a common space and distill that information to a student network. Our results highlighted increased performance and reduced bias for all our experiments. In addition, we further show that having biased/specialized teachers is crucial by showing that our approach achieves better results than when knowledge is distilled from four teachers trained on balanced datasets. Our approach represents a step forward to the understanding of the importance of ethnicity-specific features.
We introduce LLaVA-MoD, a novel framework designed to enable the efficient training of small-scale Multimodal Language Models (s-MLLM) by distilling knowledge from large-scale MLLM (l-MLLM). Our approach tackles two fundamental challenges in MLLM distillation. First, we optimize the network structure of s-MLLM by integrating a sparse Mixture of Experts (MoE) architecture into the language model, striking a balance between computational efficiency and model expressiveness. Second, we propose a progressive knowledge transfer strategy to ensure comprehensive knowledge migration. This strategy begins with mimic distillation, where we minimize the Kullback-Leibler (KL) divergence between output distributions to enable the student model to emulate the teacher network's understanding. Following this, we introduce preference distillation via Direct Preference Optimization (DPO), where the key lies in treating l-MLLM as the reference model. During this phase, the s-MLLM's ability to discriminate between superior and inferior examples is significantly enhanced beyond l-MLLM, leading to a better student that surpasses its teacher, particularly in hallucination benchmarks. Extensive experiments demonstrate that LLaVA-MoD outperforms existing models across various multimodal benchmarks while maintaining a minimal number of activated parameters and low computational costs. Remarkably, LLaVA-MoD, with only 2B activated parameters, surpasses Qwen-VL-Chat-7B by an average of 8.8% across benchmarks, using merely 0.3% of the training data and 23% trainable parameters. These results underscore LLaVA-MoD's ability to effectively distill comprehensive knowledge from its teacher model, paving the way for the development of more efficient MLLMs. The code will be available on: https://github.com/shufangxun/LLaVA-MoD.
Recent studies showcase the competitive accuracy of Vision Transformers (ViTs) in relation to Convolutional Neural Networks (CNNs), along with their remarkable robustness. However, ViTs demand a large amount of data to achieve adequate performance, which makes their application to small datasets challenging, falling behind CNNs. To overcome this, we propose GenFormer, a data augmentation strategy utilizing generated images, thereby improving transformer accuracy and robustness on small-scale image classification tasks. In our comprehensive evaluation we propose Tiny ImageNetV2, -R, and -A as new test set variants of Tiny ImageNet by transferring established ImageNet generalization and robustness benchmarks to the small-scale data domain. Similarly, we introduce MedMNIST-C and EuroSAT-C as corrupted test set variants of established fine-grained datasets in the medical and aerial domain. Through a series of experiments conducted on small datasets of various domains, including Tiny ImageNet, CIFAR, EuroSAT and MedMNIST datasets, we demonstrate the synergistic power of our method, in particular when combined with common train and test time augmentations, knowledge distillation, and architectural design choices. Additionally, we prove the effectiveness of our approach under challenging conditions with limited training data, demonstrating significant improvements in both accuracy and robustness, bridging the gap between CNNs and ViTs in the small-scale dataset domain.
Large Language Models (LLMs) have emerged as promising solutions for a variety of medical and clinical decision support applications. However, LLMs are often subject to different types of biases, which can lead to unfair treatment of individuals, worsening health disparities, and reducing trust in AI-augmented medical tools. Aiming to address this important issue, in this study, we present a new model alignment approach for aligning LLMs using a preference optimization method within a knowledge distillation framework. Prior to presenting our proposed method, we first use an evaluation framework to conduct a comprehensive (largest to our knowledge) empirical evaluation to reveal the type and nature of existing biases in LLMs used for medical applications. We then offer a bias mitigation technique to reduce the unfair patterns in LLM outputs across different subgroups identified by the protected attributes. We show that our mitigation method is effective in significantly reducing observed biased patterns. Our code is publicly available at \url{https://github.com/healthylaife/FairAlignmentLLM}.
Human emotion is a complex phenomenon conveyed and perceived through facial expressions, vocal tones, body language, and physiological signals. Multimodal emotion recognition systems can perform well because they can learn complementary and redundant semantic information from diverse sensors. In real-world scenarios, only a subset of the modalities employed for training may be available at test time. Learning privileged information allows a model to exploit data from additional modalities that are only available during training. SOTA methods for PKD have been proposed to distill information from a teacher model (with privileged modalities) to a student model (without privileged modalities). However, such PKD methods utilize point-to-point matching and do not explicitly capture the relational information. Recently, methods have been proposed to distill the structural information. However, PKD methods based on structural similarity are primarily confined to learning from a single joint teacher representation, which limits their robustness, accuracy, and ability to learn from diverse multimodal sources. In this paper, a multi-teacher PKD (MT-PKDOT) method with self-distillation is introduced to align diverse teacher representations before distilling them to the student. MT-PKDOT employs a structural similarity KD mechanism based on a regularized optimal transport (OT) for distillation. The proposed MT-PKDOT method was validated on the Affwild2 and Biovid datasets. Results indicate that our proposed method can outperform SOTA PKD methods. It improves the visual-only baseline on Biovid data by 5.5%. On the Affwild2 dataset, the proposed method improves 3% and 5% over the visual-only baseline for valence and arousal respectively. Allowing the student to learn from multiple diverse sources is shown to increase the accuracy and implicitly avoids negative transfer to the student model.
Event cameras, known for low-latency operation and superior performance in challenging lighting conditions, are suitable for sensitive computer vision tasks such as semantic segmentation in autonomous driving. However, challenges arise due to limited event-based data and the absence of large-scale segmentation benchmarks. Current works are confined to closed-set semantic segmentation, limiting their adaptability to other applications. In this paper, we introduce OVOSE, the first Open-Vocabulary Semantic Segmentation algorithm for Event cameras. OVOSE leverages synthetic event data and knowledge distillation from a pre-trained image-based foundation model to an event-based counterpart, effectively preserving spatial context and transferring open-vocabulary semantic segmentation capabilities. We evaluate the performance of OVOSE on two driving semantic segmentation datasets DDD17, and DSEC-Semantic, comparing it with existing conventional image open-vocabulary models adapted for event-based data. Similarly, we compare OVOSE with state-of-the-art methods designed for closed-set settings in unsupervised domain adaptation for event-based semantic segmentation. OVOSE demonstrates superior performance, showcasing its potential for real-world applications. The code is available at https://github.com/ram95d/OVOSE.
Recent research on knowledge distillation has increasingly focused on logit distillation because of its simplicity, effectiveness, and versatility in model compression. In this paper, we introduce Refined Logit Distillation (RLD) to address the limitations of current logit distillation methods. Our approach is motivated by the observation that even high-performing teacher models can make incorrect predictions, creating a conflict between the standard distillation loss and the cross-entropy loss. This conflict can undermine the consistency of the student model's learning objectives. Previous attempts to use labels to empirically correct teacher predictions may undermine the class correlation. In contrast, our RLD employs labeling information to dynamically refine teacher logits. In this way, our method can effectively eliminate misleading information from the teacher while preserving crucial class correlations, thus enhancing the value and efficiency of distilled knowledge. Experimental results on CIFAR-100 and ImageNet demonstrate its superiority over existing methods. The code is provided at \text{https://github.com/zju-SWJ/RLD}.
Video Snapshot Compressive Imaging (SCI) uses a low-speed 2D camera to capture high-speed scenes as snapshot compressed measurements, followed by a reconstruction algorithm to retrieve the high-speed video frames. The fast evolving mobile devices and existing high-performance video SCI reconstruction algorithms motivate us to develop mobile reconstruction methods for real-world applications. Yet, it is still challenging to deploy previous reconstruction algorithms on mobile devices due to the complex inference process, let alone real-time mobile reconstruction. To the best of our knowledge, there is no video SCI reconstruction model designed to run on the mobile devices. Towards this end, in this paper, we present an effective approach for video SCI reconstruction, dubbed MobileSCI, which can run at real-time speed on the mobile devices for the first time. Specifically, we first build a U-shaped 2D convolution-based architecture, which is much more efficient and mobile-friendly than previous state-of-the-art reconstruction methods. Besides, an efficient feature mixing block, based on the channel splitting and shuffling mechanisms, is introduced as a novel bottleneck block of our proposed MobileSCI to alleviate the computational burden. Finally, a customized knowledge distillation strategy is utilized to further improve the reconstruction quality. Extensive results on both simulated and real data show that our proposed MobileSCI can achieve superior reconstruction quality with high efficiency on the mobile devices. Particularly, we can reconstruct a 256 X 256 X 8 snapshot compressed measurement with real-time performance (about 35 FPS) on an iPhone 15. Code is available at https://github.com/mcao92/MobileSCI.
Diffusion-based image super-resolution (SR) methods have shown promise in reconstructing high-resolution images with fine details from low-resolution counterparts. However, these approaches typically require tens or even hundreds of iterative samplings, resulting in significant latency. Recently, techniques have been devised to enhance the sampling efficiency of diffusion-based SR models via knowledge distillation. Nonetheless, when aligning the knowledge of student and teacher models, these solutions either solely rely on pixel-level loss constraints or neglect the fact that diffusion models prioritize varying levels of information at different time steps. To accomplish effective and efficient image super-resolution, we propose a time-aware diffusion distillation method, named TAD-SR. Specifically, we introduce a novel score distillation strategy to align the data distribution between the outputs of the student and teacher models after minor noise perturbation. This distillation strategy enables the student network to concentrate more on the high-frequency details. Furthermore, to mitigate performance limitations stemming from distillation, we integrate a latent adversarial loss and devise a time-aware discriminator that leverages diffusion priors to effectively distinguish between real images and generated images. Extensive experiments conducted on synthetic and real-world datasets demonstrate that the proposed method achieves comparable or even superior performance compared to both previous state-of-the-art (SOTA) methods and the teacher model in just one sampling step. Codes are available at https://github.com/LearningHx/TAD-SR.
Many recommender models have been proposed to investigate how to incorporate multimodal content information into traditional collaborative filtering framework effectively. The use of multimodal information is expected to provide more comprehensive information and lead to superior performance. However, the integration of multiple modalities often encounters the modal imbalance problem: since the information in different modalities is unbalanced, optimizing the same objective across all modalities leads to the under-optimization problem of the weak modalities with a slower convergence rate or lower performance. Even worse, we find that in multimodal recommendation models, all modalities suffer from the problem of insufficient optimization. To address these issues, we propose a Counterfactual Knowledge Distillation method that could solve the imbalance problem and make the best use of all modalities. Through modality-specific knowledge distillation, it could guide the multimodal model to learn modality-specific knowledge from uni-modal teachers. We also design a novel generic-and-specific distillation loss to guide the multimodal student to learn wider-and-deeper knowledge from teachers. Additionally, to adaptively recalibrate the focus of the multimodal model towards weaker modalities during training, we estimate the causal effect of each modality on the training objective using counterfactual inference techniques, through which we could determine the weak modalities, quantify the imbalance degree and re-weight the distillation loss accordingly. Our method could serve as a plug-and-play module for both late-fusion and early-fusion backbones. Extensive experiments on six backbones show that our proposed method can improve the performance by a large margin. The source code will be released at \url{https://github.com/CRIPAC-DIG/Balanced-Multimodal-Rec}