PyTorchKR Papers

최신 논문

총 1,512편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2112.01528공식 코드 · Papers with Code

A Fast Knowledge Distillation Framework for Visual Recognition

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.

arXiv:2112.00725공식 코드 · Papers with Code

The Augmented Image Prior: Distilling 1000 Classes by Extrapolating from a Single Image

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.

arXiv:2112.00459공식 코드 · Papers with Code

Information Theoretic Representation Distillation

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

arXiv:2111.14619공식 코드 · Papers with Code

WiFi-based Multi-task Sensing

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.

arXiv:2111.12341공식 코드 · Papers with Code

EvDistill: Asynchronous Events to End-task Learning via Bidirectional Reconstruction-guided Cross-modal Knowledge Distillation

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.

arXiv:2111.12624공식 코드 · Papers with Code

Self-slimmed Vision Transformer

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.

arXiv:2111.11747공식 코드 · Papers with Code

Semi-Online Knowledge Distillation

Knowledge distillation is an effective and stable method for model compression via knowledge transfer. Conventional knowledge distillation (KD) is to transfer knowledge from a large and well pre-trained teacher network to a small student network, which is a one-way process. Recently, deep mutual learning (DML) has been proposed to help student networks learn collaboratively and simultaneously. However, to the best of our knowledge, KD and DML have never been jointly explored in a unified framework to solve the knowledge distillation problem. In this paper, we investigate that the teacher model supports more trustworthy supervision signals in KD, while the student captures more similar behaviors from the teacher in DML. Based on these observations, we first propose to combine KD with DML in a unified framework. Furthermore, we propose a Semi-Online Knowledge Distillation (SOKD) method that effectively improves the performance of the student and the teacher. In this method, we introduce the peer-teaching training fashion in DML in order to alleviate the student's imitation difficulty, and also leverage the supervision signals provided by the well-trained teacher in KD. Besides, we also show our framework can be easily extended to feature-based distillation methods. Extensive experiments on CIFAR-100 and ImageNet datasets demonstrate the proposed method achieves state-of-the-art performance.

arXiv:2111.11837공식 코드 · Papers with Code

Focal and Global Knowledge Distillation for Detectors

Knowledge distillation has been applied to image classification successfully. However, object detection is much more sophisticated and most knowledge distillation methods have failed on it. In this paper, we point out that in object detection, the features of the teacher and student vary greatly in different areas, especially in the foreground and background. If we distill them equally, the uneven differences between feature maps will negatively affect the distillation. Thus, we propose Focal and Global Distillation (FGD). Focal distillation separates the foreground and background, forcing the student to focus on the teacher's critical pixels and channels. Global distillation rebuilds the relation between different pixels and transfers it from teachers to students, compensating for missing global information in focal distillation. As our method only needs to calculate the loss on the feature map, FGD can be applied to various detectors. We experiment on various detectors with different backbones and the results show that the student detector achieves excellent mAP improvement. For example, ResNet-50 based RetinaNet, Faster RCNN, RepPoints and Mask RCNN with our distillation method achieve 40.7%, 42.0%, 42.0% and 42.1% mAP on COCO2017, which are 3.3, 3.6, 3.4 and 2.9 higher than the baseline, respectively. Our codes are available at https://github.com/yzd-v/FGD.

arXiv:2111.07047공식 코드 · Papers with Code

Facial Landmark Points Detection Using Knowledge Distillation-Based Neural Networks

Facial landmark detection is a vital step for numerous facial image analysis applications. Although some deep learning-based methods have achieved good performances in this task, they are often not suitable for running on mobile devices. Such methods rely on networks with many parameters, which makes the training and inference time-consuming. Training lightweight neural networks such as MobileNets are often challenging, and the models might have low accuracy. Inspired by knowledge distillation (KD), this paper presents a novel loss function to train a lightweight Student network (e.g., MobileNetV2) for facial landmark detection. We use two Teacher networks, a Tolerant-Teacher and a Tough-Teacher in conjunction with the Student network. The Tolerant-Teacher is trained using Soft-landmarks created by active shape models, while the Tough-Teacher is trained using the ground truth (aka Hard-landmarks) landmark points. To utilize the facial landmark points predicted by the Teacher networks, we define an Assistive Loss (ALoss) for each Teacher network. Moreover, we define a loss function called KD-Loss that utilizes the facial landmark points predicted by the two pre-trained Teacher networks (EfficientNet-b3) to guide the lightweight Student network towards predicting the Hard-landmarks. Our experimental results on three challenging facial datasets show that the proposed architecture will result in a better-trained Student network that can extract facial landmark points with high accuracy.

arXiv:2111.07239공식 코드 · Papers with Code

Robust and Accurate Object Detection via Self-Knowledge Distillation

Object detection has achieved promising performance on clean datasets, but how to achieve better tradeoff between the adversarial robustness and clean precision is still under-explored. Adversarial training is the mainstream method to improve robustness, but most of the works will sacrifice clean precision to gain robustness than standard training. In this paper, we propose Unified Decoupled Feature Alignment (UDFA), a novel fine-tuning paradigm which achieves better performance than existing methods, by fully exploring the combination between self-knowledge distillation and adversarial training for object detection. We first use decoupled fore/back-ground features to construct self-knowledge distillation branch between clean feature representation from pretrained detector (served as teacher) and adversarial feature representation from student detector. Then we explore the self-knowledge distillation from a new angle by decoupling original branch into a self-supervised learning branch and a new self-knowledge distillation branch. With extensive experiments on the PASCAL-VOC and MS-COCO benchmarks, the evaluation results show that UDFA can surpass the standard training and state-of-the-art adversarial training methods for object detection. For example, compared with teacher detector, our approach on GFLV2 with ResNet-50 improves clean precision by 2.2 AP on PASCAL-VOC; compared with SOTA adversarial training methods, our approach improves clean precision by 1.6 AP, while improving adversarial robustness by 0.5 AP. Our code will be available at https://github.com/grispeut/udfa.

arXiv:2111.04993공식 코드 · Papers with Code

Incremental Meta-Learning via Episodic Replay Distillation for Few-Shot Image Recognition

Most meta-learning approaches assume the existence of a very large set of labeled data available for episodic meta-learning of base knowledge. This contrasts with the more realistic continual learning paradigm in which data arrives incrementally in the form of tasks containing disjoint classes. In this paper we consider this problem of Incremental Meta-Learning (IML) in which classes are presented incrementally in discrete tasks. We propose an approach to IML, which we call Episodic Replay Distillation (ERD), that mixes classes from the current task with class exemplars from previous tasks when sampling episodes for meta-learning. These episodes are then used for knowledge distillation to minimize catastrophic forgetting. Experiments on four datasets demonstrate that ERD surpasses the state-of-the-art. In particular, on the more challenging one-shot, long task sequence incremental meta-learning scenarios, we reduce the gap between IML and the joint-training upper bound from 3.5% / 10.1% / 13.4% with the current state-of-the-art to 2.6% / 2.9% / 5.0% with our method on Tiered-ImageNet / Mini-ImageNet / CIFAR100, respectively.

arXiv:2111.04964공식 코드 · Papers with Code

On Representation Knowledge Distillation for Graph Neural Networks

Knowledge distillation is a learning paradigm for boosting resource-efficient graph neural networks (GNNs) using more expressive yet cumbersome teacher models. Past work on distillation for GNNs proposed the Local Structure Preserving loss (LSP), which matches local structural relationships defined over edges across the student and teacher's node embeddings. This paper studies whether preserving the global topology of how the teacher embeds graph data can be a more effective distillation objective for GNNs, as real-world graphs often contain latent interactions and noisy edges. We propose Graph Contrastive Representation Distillation (G-CRD), which uses contrastive learning to implicitly preserve global topology by aligning the student node embeddings to those of the teacher in a shared representation space. Additionally, we introduce an expanded set of benchmarks on large-scale real-world datasets where the performance gap between teacher and student GNNs is non-negligible. Experiments across 4 datasets and 14 heterogeneous GNN architectures show that G-CRD consistently boosts the performance and robustness of lightweight GNNs, outperforming LSP (and a global structure preserving variant of LSP) as well as baselines from 2D computer vision. An analysis of the representational similarity among teacher and student embedding spaces reveals that G-CRD balances preserving local and global relationships, while structure preserving approaches are best at preserving one or the other. Our code is available at https://github.com/chaitjo/efficient-gnns

arXiv:2111.00775공식 코드 · Papers with Code

PP-ShiTu: A Practical Lightweight Image Recognition System

In recent years, image recognition applications have developed rapidly. A large number of studies and techniques have emerged in different fields, such as face recognition, pedestrian and vehicle re-identification, landmark retrieval, and product recognition. In this paper, we propose a practical lightweight image recognition system, named PP-ShiTu, consisting of the following 3 modules, mainbody detection, feature extraction and vector search. We introduce popular strategies including metric learning, deep hash, knowledge distillation and model quantization to improve accuracy and inference speed. With strategies above, PP-ShiTu works well in different scenarios with a set of models trained on a mixed dataset. Experiments on different datasets and benchmarks show that the system is widely effective in different domains of image recognition. All the above mentioned models are open-sourced and the code is available in the GitHub repository PaddleClas on PaddlePaddle.

arXiv:2111.00674공식 코드 · Papers with Code

Distilling Object Detectors with Feature Richness

In recent years, large-scale deep models have achieved great success, but the huge computational complexity and massive storage requirements make it a great challenge to deploy them in resource-limited devices. As a model compression and acceleration method, knowledge distillation effectively improves the performance of small models by transferring the dark knowledge from the teacher detector. However, most of the existing distillation-based detection methods mainly imitating features near bounding boxes, which suffer from two limitations. First, they ignore the beneficial features outside the bounding boxes. Second, these methods imitate some features which are mistakenly regarded as the background by the teacher detector. To address the above issues, we propose a novel Feature-Richness Score (FRS) method to choose important features that improve generalized detectability during distilling. The proposed method effectively retrieves the important features outside the bounding boxes and removes the detrimental features within the bounding boxes. Extensive experiments show that our methods achieve excellent performance on both anchor-based and anchor-free detectors. For example, RetinaNet with ResNet-50 achieves 39.7% in mAP on the COCO2017 dataset, which even surpasses the ResNet-101 based teacher detector 38.9% by 0.8%. Our implementation is available at https://github.com/duzhixing/FRS.

arXiv:2111.00643공식 코드 · Papers with Code

Learning Distilled Collaboration Graph for Multi-Agent Perception

To promote better performance-bandwidth trade-off for multi-agent perception, we propose a novel distilled collaboration graph (DiscoGraph) to model trainable, pose-aware, and adaptive collaboration among agents. Our key novelties lie in two aspects. First, we propose a teacher-student framework to train DiscoGraph via knowledge distillation. The teacher model employs an early collaboration with holistic-view inputs; the student model is based on intermediate collaboration with single-view inputs. Our framework trains DiscoGraph by constraining post-collaboration feature maps in the student model to match the correspondences in the teacher model. Second, we propose a matrix-valued edge weight in DiscoGraph. In such a matrix, each element reflects the inter-agent attention at a specific spatial region, allowing an agent to adaptively highlight the informative regions. During inference, we only need to use the student model named as the distilled collaboration network (DiscoNet). Attributed to the teacher-student framework, multiple agents with the shared DiscoNet could collaboratively approach the performance of a hypothetical teacher model with a holistic view. Our approach is validated on V2X-Sim 1.0, a large-scale multi-agent perception dataset that we synthesized using CARLA and SUMO co-simulation. Our quantitative and qualitative experiments in multi-agent 3D object detection show that DiscoNet could not only achieve a better performance-bandwidth trade-off than the state-of-the-art collaborative perception methods, but also bring more straightforward design rationale. Our code is available on https://github.com/ai4ce/DiscoNet.

arXiv:2110.15094공식 코드 · Papers with Code

Mosaicking to Distill: Knowledge Distillation from Out-of-Domain Data

Knowledge distillation~(KD) aims to craft a compact student model that imitates the behavior of a pre-trained teacher in a target domain. Prior KD approaches, despite their gratifying results, have largely relied on the premise that \emph{in-domain} data is available to carry out the knowledge transfer. Such an assumption, unfortunately, in many cases violates the practical setting, since the original training data or even the data domain is often unreachable due to privacy or copyright reasons. In this paper, we attempt to tackle an ambitious task, termed as \emph{out-of-domain} knowledge distillation~(OOD-KD), which allows us to conduct KD using only OOD data that can be readily obtained at a very low cost. Admittedly, OOD-KD is by nature a highly challenging task due to the agnostic domain gap. To this end, we introduce a handy yet surprisingly efficacious approach, dubbed as~\textit{MosaicKD}. The key insight behind MosaicKD lies in that, samples from various domains share common local patterns, even though their global semantic may vary significantly; these shared local patterns, in turn, can be re-assembled analogous to mosaic tiling, to approximate the in-domain data and to further alleviating the domain discrepancy. In MosaicKD, this is achieved through a four-player min-max game, in which a generator, a discriminator, a student network, are collectively trained in an adversarial manner, partially under the guidance of a pre-trained teacher. We validate MosaicKD over {classification and semantic segmentation tasks} across various benchmarks, and demonstrate that it yields results much superior to the state-of-the-art counterparts on OOD data. Our code is available at \url{https://github.com/zju-vipa/MosaicKD}.

arXiv:2110.14553공식 코드 · Papers with Code

GenURL: A General Framework for Unsupervised Representation Learning

Unsupervised representation learning (URL), which learns compact embeddings of high-dimensional data without supervision, has made remarkable progress recently. However, the development of URLs for different requirements is independent, which limits the generalization of the algorithms, especially prohibitive as the number of tasks grows. For example, dimension reduction methods, t-SNE, and UMAP optimize pair-wise data relationships by preserving the global geometric structure, while self-supervised learning, SimCLR, and BYOL focus on mining the local statistics of instances under specific augmentations. To address this dilemma, we summarize and propose a unified similarity-based URL framework, GenURL, which can smoothly adapt to various URL tasks. In this paper, we regard URL tasks as different implicit constraints on the data geometric structure that help to seek optimal low-dimensional representations that boil down to data structural modeling (DSM) and low-dimensional transformation (LDT). Specifically, DMS provides a structure-based submodule to describe the global structures, and LDT learns compact low-dimensional embeddings with given pretext tasks. Moreover, an objective function, General Kullback-Leibler divergence (GKL), is proposed to connect DMS and LDT naturally. Comprehensive experiments demonstrate that GenURL achieves consistent state-of-the-art performance in self-supervised visual learning, unsupervised knowledge distillation (KD), graph embeddings (GE), and dimension reduction.

arXiv:2110.12724공식 코드 · Papers with Code

Instance-Conditional Knowledge Distillation for Object Detection

Knowledge distillation has shown great success in classification, however, it is still challenging for detection. In a typical image for detection, representations from different locations may have different contributions to detection targets, making the distillation hard to balance. In this paper, we propose a conditional distillation framework to distill the desired knowledge, namely knowledge that is beneficial in terms of both classification and localization for every instance. The framework introduces a learnable conditional decoding module, which retrieves information given each target instance as query. Specifically, we encode the condition information as query and use the teacher's representations as key. The attention between query and key is used to measure the contribution of different features, guided by a localization-recognition-sensitive auxiliary task. Extensive experiments demonstrate the efficacy of our method: we observe impressive improvements under various settings. Notably, we boost RetinaNet with ResNet-50 backbone from 37.4 to 40.7 mAP (+3.3) under 1x schedule, that even surpasses the teacher (40.4 mAP) with ResNet-101 backbone under 3x schedule. Code has been released on https://github.com/megvii-research/ICD.

arXiv:2110.11650공식 코드 · Papers with Code

Pixel-by-Pixel Cross-Domain Alignment for Few-Shot Semantic Segmentation

In this paper we consider the task of semantic segmentation in autonomous driving applications. Specifically, we consider the cross-domain few-shot setting where training can use only few real-world annotated images and many annotated synthetic images. In this context, aligning the domains is made more challenging by the pixel-wise class imbalance that is intrinsic in the segmentation and that leads to ignoring the underrepresented classes and overfitting the well represented ones. We address this problem with a novel framework called Pixel-By-Pixel Cross-Domain Alignment (PixDA). We propose a novel pixel-by-pixel domain adversarial loss following three criteria: (i) align the source and the target domain for each pixel, (ii) avoid negative transfer on the correctly represented pixels, and (iii) regularize the training of infrequent classes to avoid overfitting. The pixel-wise adversarial training is assisted by a novel sample selection procedure, that handles the imbalance between source and target data, and a knowledge distillation strategy, that avoids overfitting towards the few target images. We demonstrate on standard synthetic-to-real benchmarks that PixDA outperforms previous state-of-the-art methods in (1-5)-shot settings.

arXiv:2110.09910공식 코드 · Papers with Code

FedHe: Heterogeneous Models and Communication-Efficient Federated Learning

Federated learning (FL) is able to manage edge devices to cooperatively train a model while maintaining the training data local and private. One common assumption in FL is that all edge devices share the same machine learning model in training, for example, identical neural network architecture. However, the computation and store capability of different devices may not be the same. Moreover, reducing communication overheads can improve the training efficiency though it is still a challenging problem in FL. In this paper, we propose a novel FL method, called FedHe, inspired by knowledge distillation, which can train heterogeneous models and support asynchronous training processes with significantly reduced communication overheads. Our analysis and experimental results demonstrate that the performance of our proposed method is better than the state-of-the-art algorithms in terms of communication overheads and model accuracy.