Despite significant advancements of deep learning-based forgery detectors for distinguishing manipulated deepfake images, most detection approaches suffer from moderate to significant performance degradation with low-quality compressed deepfake images. Because of the limited information in low-quality images, detecting low-quality deepfake remains an important challenge. In this work, we apply frequency domain learning and optimal transport theory in knowledge distillation (KD) to specifically improve the detection of low-quality compressed deepfake images. We explore transfer learning capability in KD to enable a student network to learn discriminative features from low-quality images effectively. In particular, we propose the Attention-based Deepfake detection Distiller (ADD), which consists of two novel distillations: 1) frequency attention distillation that effectively retrieves the removed high-frequency components in the student network, and 2) multi-view attention distillation that creates multiple attention vectors by slicing the teacher's and student's tensors under different views to transfer the teacher tensor's distribution to the student more efficiently. Our extensive experimental results demonstrate that our approach outperforms state-of-the-art baselines in detecting low-quality compressed deepfake images.
Incremental learning represents a crucial task in aerial image processing, especially given the limited availability of large-scale annotated datasets. A major issue concerning current deep neural architectures is known as catastrophic forgetting, namely the inability to faithfully maintain past knowledge once a new set of data is provided for retraining. Over the years, several techniques have been proposed to mitigate this problem for image classification and object detection. However, only recently the focus has shifted towards more complex downstream tasks such as instance or semantic segmentation. Starting from incremental-class learning for semantic segmentation tasks, our goal is to adapt this strategy to the aerial domain, exploiting a peculiar feature that differentiates it from natural images, namely the orientation. In addition to the standard knowledge distillation approach, we propose a contrastive regularization, where any given input is compared with its augmented version (i.e. flipping and rotations) in order to minimize the difference between the segmentation features produced by both inputs. We show the effectiveness of our solution on the Potsdam dataset, outperforming the incremental baseline in every test. Code available at: https://github.com/edornd/contrastive-distillation.
Knowledge distillation (KD) has recently emerged as a powerful strategy to transfer knowledge from a pre-trained teacher model to a lightweight student, and has demonstrated its unprecedented success over a wide spectrum of applications. In spite of the encouraging results, the KD process per se poses a potential threat to network ownership protection, since the knowledge contained in network can be effortlessly distilled and hence exposed to a malicious user. In this paper, we propose a novel framework, termed as Safe Distillation Box (SDB), that allows us to wrap a pre-trained model in a virtual box for intellectual property protection. Specifically, SDB preserves the inference capability of the wrapped model to all users, but precludes KD from unauthorized users. For authorized users, on the other hand, SDB carries out a knowledge augmentation scheme to strengthen the KD performances and the results of the student model. In other words, all users may employ a model in SDB for inference, but only authorized users get access to KD from the model. The proposed SDB imposes no constraints over the model architecture, and may readily serve as a plug-and-play solution to protect the ownership of a pre-trained network. Experiments across various datasets and architectures demonstrate that, with SDB, the performance of an unauthorized KD drops significantly while that of an authorized gets enhanced, demonstrating the effectiveness of SDB.
Time-series classification approaches based on deep neural networks are easy to be overfitting on UCR datasets, which is caused by the few-shot problem of those datasets. Therefore, in order to alleviate the overfitting phenomenon for further improving the accuracy, we first propose Label Smoothing for InceptionTime (LSTime), which adopts the information of soft labels compared to just hard labels. Next, instead of manually adjusting soft labels by LSTime, Knowledge Distillation for InceptionTime (KDTime) is proposed in order to automatically generate soft labels by the teacher model. At last, in order to rectify the incorrect predicted soft labels from the teacher model, Knowledge Distillation with Calibration for InceptionTime (KDCTime) is proposed, where it contains two optional calibrating strategies, i.e. KDC by Translating (KDCT) and KDC by Reordering (KDCR). The experimental results show that the accuracy of KDCTime is promising, while its inference time is two orders of magnitude faster than ROCKET with an acceptable training time overhead.
This paper studies continual learning (CL) of a sequence of aspect sentiment classification(ASC) tasks in a particular CL setting called domain incremental learning (DIL). Each task is from a different domain or product. The DIL setting is particularly suited to ASC because in testing the system needs not know the task/domain to which the test data belongs. To our knowledge, this setting has not been studied before for ASC. This paper proposes a novel model called CLASSIC. The key novelty is a contrastive continual learning method that enables both knowledge transfer across tasks and knowledge distillation from old tasks to the new task, which eliminates the need for task ids in testing. Experimental results show the high effectiveness of CLASSIC.
Human activity recognition (HAR) based on multi-modal approach has been recently shown to improve the accuracy performance of HAR. However, restricted computational resources associated with wearable devices, i.e., smartwatch, failed to directly support such advanced methods. To tackle this issue, this study introduces an end-to-end Vision-to-Sensor Knowledge Distillation (VSKD) framework. In this VSKD framework, only time-series data, i.e., accelerometer data, is needed from wearable devices during the testing phase. Therefore, this framework will not only reduce the computational demands on edge devices, but also produce a learning model that closely matches the performance of the computational expensive multi-modal approach. In order to retain the local temporal relationship and facilitate visual deep learning models, we first convert time-series data to two-dimensional images by applying the Gramian Angular Field ( GAF) based encoding method. We adopted ResNet18 and multi-scale TRN with BN-Inception as teacher and student network in this study, respectively. A novel loss function, named Distance and Angle-wised Semantic Knowledge loss (DASK), is proposed to mitigate the modality variations between the vision and the sensor domain. Extensive experimental results on UTD-MHAD, MMAct, and Berkeley-MHAD datasets demonstrate the effectiveness and competitiveness of the proposed VSKD model which can deployed on wearable sensors.
News recommendation is a widely adopted technique to provide personalized news feeds for the user. Recently, pre-trained language models (PLMs) have demonstrated the great capability of natural language understanding and benefited news recommendation via improving news modeling. However, most existing works simply finetune the PLM with the news recommendation task, which may suffer from the known domain shift problem between the pre-training corpus and downstream news texts. Moreover, PLMs usually contain a large volume of parameters and have high computational overhead, which imposes a great burden on low-latency online services. In this paper, we propose Tiny-NewsRec, which can improve both the effectiveness and the efficiency of PLM-based news recommendation. We first design a self-supervised domain-specific post-training method to better adapt the general PLM to the news domain with a contrastive matching task between news titles and news bodies. We further propose a two-stage knowledge distillation method to improve the efficiency of the large PLM-based news recommendation model while maintaining its performance. Multiple teacher models originated from different time steps of our post-training procedure are used to transfer comprehensive knowledge to the student in both its post-training and finetuning stage. Extensive experiments on two real-world datasets validate the effectiveness and efficiency of our method.
While Knowledge Distillation (KD) has been recognized as a useful tool in many visual tasks, such as supervised classification and self-supervised representation learning, the main drawback of a vanilla KD framework is its mechanism, which consumes the majority of the computational overhead on forwarding through the giant teacher networks, making the entire learning procedure inefficient and costly. ReLabel, a recently proposed solution, suggests creating a label map for the entire image. During training, it receives the cropped region-level label by RoI aligning on a pre-generated entire label map, allowing for efficient supervision generation without having to pass through the teachers many times. However, as the KD teachers are from conventional multi-crop training, there are various mismatches between the global label-map and region-level label in this technique, resulting in performance deterioration. In this study, we present a Fast Knowledge Distillation (FKD) framework that replicates the distillation training phase and generates soft labels using the multi-crop KD approach, while training faster than ReLabel since no post-processes such as RoI align and softmax operations are used. When conducting multi-crop in the same image for data loading, our FKD is even more efficient than the traditional image classification framework. On ImageNet-1K, we obtain 79.8% with ResNet-50, outperforming ReLabel by ~1.0% while being faster. On the self-supervised learning task, we also show that FKD has an efficiency advantage. Our project page: http://zhiqiangshen.com/projects/FKD/index.html, source code and models are available at: https://github.com/szq0214/FKD.
The robustness of federated learning (FL) is vital for the distributed training of an accurate global model that is shared among large number of clients. The collaborative learning framework by typically aggregating model updates is vulnerable to model poisoning attacks from adversarial clients. Since the shared information between the global server and participants are only limited to model parameters, it is challenging to detect bad model updates. Moreover, real-world datasets are usually heterogeneous and not independent and identically distributed (Non-IID) among participants, which makes the design of such robust FL pipeline more difficult. In this work, we propose a novel robust aggregation method, Federated Robust Adaptive Distillation (FedRAD), to detect adversaries and robustly aggregate local models based on properties of the median statistic, and then performing an adapted version of ensemble Knowledge Distillation. We run extensive experiments to evaluate the proposed method against recently published works. The results show that FedRAD outperforms all other aggregators in the presence of adversaries, as well as in heterogeneous data distributions.
What can neural networks learn about the visual world when provided with only a single image as input? While any image obviously cannot contain the multitudes of all existing objects, scenes and lighting conditions - within the space of all 256^(3x224x224) possible 224-sized square images, it might still provide a strong prior for natural images. To analyze this `augmented image prior' hypothesis, we develop a simple framework for training neural networks from scratch using a single image and augmentations using knowledge distillation from a supervised pretrained teacher. With this, we find the answer to the above question to be: `surprisingly, a lot'. In quantitative terms, we find accuracies of 94%/74% on CIFAR-10/100, 69% on ImageNet, and by extending this method to video and audio, 51% on Kinetics-400 and 84% on SpeechCommands. In extensive analyses spanning 13 datasets, we disentangle the effect of augmentations, choice of data and network architectures and also provide qualitative evaluations that include lucid `panda neurons' in networks that have never even seen one.
Despite the empirical success of knowledge distillation, current state-of-the-art methods are computationally expensive to train, which makes them difficult to adopt in practice. To address this problem, we introduce two distinct complementary losses inspired by a cheap entropy-like estimator. These losses aim to maximise the correlation and mutual information between the student and teacher representations. Our method incurs significantly less training overheads than other approaches and achieves competitive performance to the state-of-the-art on the knowledge distillation and cross-model transfer tasks. We further demonstrate the effectiveness of our method on a binary distillation task, whereby it leads to a new state-of-the-art for binary quantisation and approaches the performance of a full precision model. Code: www.github.com/roymiles/ITRD
Images posted online present a privacy concern in that they may be used as reference examples for a facial recognition system. Such abuse of images is in violation of privacy rights but is difficult to counter. It is well established that adversarial example images can be created for recognition systems which are based on deep neural networks. These adversarial examples can be used to disrupt the utility of the images as reference examples or training data. In this work we use a Generative Adversarial Network (GAN) to create adversarial examples to deceive facial recognition and we achieve an acceptable success rate in fooling the face recognition. Our results reduce the training time for the GAN by removing the discriminator component. Furthermore, our results show knowledge distillation can be employed to drastically reduce the size of the resulting model without impacting performance indicating that our contribution could run comfortably on a smartphone
WiFi-based sensing has aroused immense attention over recent years. The rationale is that the signal fluctuations caused by humans carry the information of human behavior which can be extracted from the channel state information of WiFi. Still, the prior studies mainly focus on single-task sensing (STS), e.g., gesture recognition, indoor localization, user identification. Since the fluctuations caused by gestures are highly coupling with body features and the user's location, we propose a WiFi-based multi-task sensing model (Wimuse) to perform gesture recognition, indoor localization, and user identification tasks simultaneously. However, these tasks have different difficulty levels (i.e., imbalance issue) and need task-specific information (i.e., discrepancy issue). To address these issues, the knowledge distillation technique and task-specific residual adaptor are adopted in Wimuse. We first train the STS model for each task. Then, for solving the imbalance issue, the extracted common feature in Wimuse is encouraged to get close to the counterpart features of the STS models. Further, for each task, a task-specific residual adaptor is applied to extract the task-specific compensation feature which is fused with the common feature to address the discrepancy issue. We conduct comprehensive experiments on three public datasets and evaluation suggests that Wimuse achieves state-of-the-art performance with the average accuracy of 85.20%, 98.39%, and 98.725% on the joint task of gesture recognition, indoor localization, and user identification, respectively.
As a promising distributed machine learning paradigm, Federated Learning (FL) trains a central model with decentralized data without compromising user privacy, which has made it widely used by Artificial Intelligence Internet of Things (AIoT) applications. However, the traditional FL suffers from model inaccuracy since it trains local models using hard labels of data and ignores useful information of incorrect predictions with small probabilities. Although various solutions try to tackle the bottleneck of the traditional FL, most of them introduce significant communication and memory overhead, making the deployment of large-scale AIoT devices a great challenge. To address the above problem, this paper presents a novel Distillation-based Federated Learning (DFL) architecture that enables efficient and accurate FL for AIoT applications. Inspired by Knowledge Distillation (KD) that can increase the model accuracy, our approach adds the soft targets used by KD to the FL model training, which occupies negligible network resources. The soft targets are generated by local sample predictions of each AIoT device after each round of local training and used for the next round of model training. During the local training of DFL, both soft targets and hard labels are used as approximation objectives of model predictions to improve model accuracy by supplementing the knowledge of soft targets. To further improve the performance of our DFL model, we design a dynamic adjustment strategy for tuning the ratio of two loss functions used in KD, which can maximize the use of both soft targets and hard labels. Comprehensive experimental results on well-known benchmarks show that our approach can significantly improve the model accuracy of FL with both Independent and Identically Distributed (IID) and non-IID data.
Fast stereo based 3D object detectors have made great progress recently. However, they lag far behind high-precision stereo based methods in accuracy. We argue that the main reason is due to the poor geometry-aware feature representation in 3D space. To solve this problem, we propose an efficient stereo geometry network (ESGN). The key in our ESGN is an efficient geometry-aware feature generation (EGFG) module. Our EGFG module first uses a stereo correlation and reprojection module to construct multi-scale stereo volumes in camera frustum space, second employs a multi-scale BEV projection and fusion module to generate multiple geometry-aware features. In these two steps, we adopt deep multi-scale information fusion for discriminative geometry-aware feature generation, without any complex aggregation networks. In addition, we introduce a deep geometry-aware feature distillation scheme to guide stereo feature learning with a LiDAR-based detector. The experiments are performed on the classical KITTI dataset. On KITTI test set, our ESGN outperforms the fast state-of-art-art detector YOLOStereo3D by 5.14\% on mAP$_{3d}$ at 62$ms$. To the best of our knowledge, our ESGN achieves a best trade-off between accuracy and speed. We hope that our efficient stereo geometry network can provide more possible directions for fast 3D object detection. Our source code will be released.
Knowledge distillation has become an important approach to obtain a compact yet effective model. To achieve this goal, a small student model is trained to exploit the knowledge of a large well-trained teacher model. However, due to the capacity gap between the teacher and the student, the student's performance is hard to reach the level of the teacher. Regarding this issue, existing methods propose to reduce the difficulty of the teacher's knowledge via a proxy way. We argue that these proxy-based methods overlook the knowledge loss of the teacher, which may cause the student to encounter capacity bottlenecks. In this paper, we alleviate the capacity gap problem from a new perspective with the purpose of averting knowledge loss. Instead of sacrificing part of the teacher's knowledge, we propose to build a more powerful student via adversarial collaborative learning. To this end, we further propose an Adversarial Collaborative Knowledge Distillation (ACKD) method that effectively improves the performance of knowledge distillation. Specifically, we construct the student model with multiple auxiliary learners. Meanwhile, we devise an adversarial collaborative module (ACM) that introduces attention mechanism and adversarial learning to enhance the capacity of the student. Extensive experiments on four classification tasks show the superiority of the proposed ACKD.
This paper describes our submission to the constrained track of WMT21 shared news translation task. We focus on the three relatively low resource language pairs Bengali to and from Hindi, English to and from Hausa, and Xhosa to and from Zulu. To overcome the limitation of relatively low parallel data we train a multilingual model using a multitask objective employing both parallel and monolingual data. In addition, we augment the data using back translation. We also train a bilingual model incorporating back translation and knowledge distillation then combine the two models using sequence-to-sequence mapping. We see around 70% relative gain in BLEU point for English to and from Hausa, and around 25% relative improvements for both Bengali to and from Hindi, and Xhosa to and from Zulu compared to bilingual baselines.
Event cameras sense per-pixel intensity changes and produce asynchronous event streams with high dynamic range and less motion blur, showing advantages over conventional cameras. A hurdle of training event-based models is the lack of large qualitative labeled data. Prior works learning end-tasks mostly rely on labeled or pseudo-labeled datasets obtained from the active pixel sensor (APS) frames; however, such datasets' quality is far from rivaling those based on the canonical images. In this paper, we propose a novel approach, called \textbf{EvDistill}, to learn a student network on the unlabeled and unpaired event data (target modality) via knowledge distillation (KD) from a teacher network trained with large-scale, labeled image data (source modality). To enable KD across the unpaired modalities, we first propose a bidirectional modality reconstruction (BMR) module to bridge both modalities and simultaneously exploit them to distill knowledge via the crafted pairs, causing no extra computation in the inference. The BMR is improved by the end-tasks and KD losses in an end-to-end manner. Second, we leverage the structural similarities of both modalities and adapt the knowledge by matching their distributions. Moreover, as most prior feature KD methods are uni-modality and less applicable to our problem, we propose to leverage an affinity graph KD loss to boost the distillation. Our extensive experiments on semantic segmentation and object recognition demonstrate that EvDistill achieves significantly better results than the prior works and KD with only events and APS frames.
Vision transformers (ViTs) have become the popular structures and outperformed convolutional neural networks (CNNs) on various vision tasks. However, such powerful transformers bring a huge computation burden, because of the exhausting token-to-token comparison. The previous works focus on dropping insignificant tokens to reduce the computational cost of ViTs. But when the dropping ratio increases, this hard manner will inevitably discard the vital tokens, which limits its efficiency. To solve the issue, we propose a generic self-slimmed learning approach for vanilla ViTs, namely SiT. Specifically, we first design a novel Token Slimming Module (TSM), which can boost the inference efficiency of ViTs by dynamic token aggregation. As a general method of token hard dropping, our TSM softly integrates redundant tokens into fewer informative ones. It can dynamically zoom visual attention without cutting off discriminative token relations in the images, even with a high slimming ratio. Furthermore, we introduce a concise Feature Recalibration Distillation (FRD) framework, wherein we design a reverse version of TSM (RTSM) to recalibrate the unstructured token in a flexible auto-encoder manner. Due to the similar structure between teacher and student, our FRD can effectively leverage structure knowledge for better convergence. Finally, we conduct extensive experiments to evaluate our SiT. It demonstrates that our method can speed up ViTs by 1.7x with negligible accuracy drop, and even speed up ViTs by 3.6x while maintaining 97% of their performance. Surprisingly, by simply arming LV-ViT with our SiT, we achieve new state-of-the-art performance on ImageNet. Code is available at https://github.com/Sense-X/SiT.
Recent advancements in self-supervised learning have reduced the gap between supervised and unsupervised representation learning. However, most self-supervised and deep clustering techniques rely heavily on data augmentation, rendering them ineffective for many learning tasks where insufficient domain knowledge exists for performing augmentation. We propose a new self-distillation based algorithm for domain-agnostic clustering. Our method builds upon the existing deep clustering frameworks and requires no separate student model. The proposed method outperforms existing domain agnostic (augmentation-free) algorithms on CIFAR-10. We empirically demonstrate that knowledge distillation can improve unsupervised representation learning by extracting richer `dark knowledge' from the model than using predicted labels alone. Preliminary experiments also suggest that self-distillation improves the convergence of DeepCluster-v2.