PyTorchKR Papers

최신 논문

총 4,526편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2203.05103

Improving Neural ODEs via Knowledge Distillation

Neural Ordinary Differential Equations (Neural ODEs) construct the continuous dynamics of hidden units using ordinary differential equations specified by a neural network, demonstrating promising results on many tasks. However, Neural ODEs still do not perform well on image recognition tasks. The possible reason is that the one-hot encoding vector commonly used in Neural ODEs can not provide enough supervised information. We propose a new training based on knowledge distillation to construct more powerful and robust Neural ODEs fitting image recognition tasks. Specially, we model the training of Neural ODEs into a teacher-student learning process, in which we propose ResNets as the teacher model to provide richer supervised information. The experimental results show that the new training manner can improve the classification accuracy of Neural ODEs by 24% on CIFAR10 and 5% on SVHN. In addition, we also quantitatively discuss the effect of both knowledge distillation and time horizon in Neural ODEs on robustness against adversarial examples. The experimental analysis concludes that introducing the knowledge distillation and increasing the time horizon can improve the robustness of Neural ODEs against adversarial examples.

arXiv:2112.09786

Distill and De-bias: Mitigating Bias in Face Verification using Knowledge Distillation

Face recognition networks generally demonstrate bias with respect to sensitive attributes like gender, skintone etc. For gender and skintone, we observe that the regions of the face that a network attends to vary by the category of an attribute. This might contribute to bias. Building on this intuition, we propose a novel distillation-based approach called Distill and De-bias (D&D) to enforce a network to attend to similar face regions, irrespective of the attribute category. In D&D, we train a teacher network on images from one category of an attribute; e.g. light skintone. Then distilling information from the teacher, we train a student network on images of the remaining category; e.g., dark skintone. A feature-level distillation loss constrains the student network to generate teacher-like representations. This allows the student network to attend to similar face regions for all attribute categories and enables it to reduce bias. We also propose a second distillation step on top of D&D, called D&D++. Here, we distill the `un-biasedness' of the D&D network into a new student network, the D&D++ network, while training this new network on all attribute categories; e.g., both light and dark skintones. This helps us train a network that is less biased for an attribute, while obtaining higher face verification performance than D&D. We show that D&D++ outperforms existing baselines in reducing gender and skintone bias on the IJB-C dataset, while obtaining higher face verification performance than existing adversarial de-biasing methods. We evaluate the effectiveness of our proposed methods on two state-of-the-art face recognition networks: ArcFace and Crystalface.

arXiv:2203.04825공식 코드 · Papers with Code

Efficient Sub-structured Knowledge Distillation

Structured prediction models aim at solving a type of problem where the output is a complex structure, rather than a single variable. Performing knowledge distillation for such models is not trivial due to their exponentially large output space. In this work, we propose an approach that is much simpler in its formulation and far more efficient for training than existing approaches. Specifically, we transfer the knowledge from a teacher model to its student model by locally matching their predictions on all sub-structures, instead of the whole output space. In this manner, we avoid adopting some time-consuming techniques like dynamic programming (DP) for decoding output structures, which permits parallel computation and makes the training process even faster in practice. Besides, it encourages the student model to better mimic the internal behavior of the teacher model. Experiments on two structured prediction tasks demonstrate that our approach outperforms previous methods and halves the time cost for one training epoch.

arXiv:2203.04314공식 코드 · Papers with Code

PyNET-QxQ: An Efficient PyNET Variant for QxQ Bayer Pattern Demosaicing in CMOS Image Sensors

Deep learning-based image signal processor (ISP) models for mobile cameras can generate high-quality images that rival those of professional DSLR cameras. However, their computational demands often make them unsuitable for mobile settings. Additionally, modern mobile cameras employ non-Bayer color filter arrays (CFA) such as Quad Bayer, Nona Bayer, and QxQ Bayer to enhance image quality, yet most existing deep learning-based ISP (or demosaicing) models focus primarily on standard Bayer CFAs. In this study, we present PyNET-QxQ, a lightweight demosaicing model specifically designed for QxQ Bayer CFA patterns, which is derived from the original PyNET. We also propose a knowledge distillation method called progressive distillation to train the reduced network more effectively. Consequently, PyNET-QxQ contains less than 2.5% of the parameters of the original PyNET while preserving its performance. Experiments using QxQ images captured by a proto type QxQ camera sensor show that PyNET-QxQ outperforms existing conventional algorithms in terms of texture and edge reconstruction, despite its significantly reduced parameter count.

arXiv:2203.04300

Multi-trial Neural Architecture Search with Lottery Tickets

Neural architecture search (NAS) has brought significant progress in recent image recognition tasks. Most existing NAS methods apply restricted search spaces, which limits the upper-bound performance of searched models. To address this issue, we propose a new search space named MobileNet3-MT. By reducing human-prior knowledge in omni dimensions of networks, MobileNet3-MT accommodates more potential candidates. For searching in this challenging search space, we present an efficient Multi-trial Evolution-based NAS method termed MENAS. Specifically, we accelerate the evolutionary search process by gradually pruning models in the population. Each model is trained with an early stop and replaced by its Lottery Tickets (the explored optimal pruned network).In this way, the full training pipeline of cumbersome networks is prevented and more efficient networks are automatically generated. Extensive experimental results on ImageNet-1K, CIFAR-10, and CIFAR-100 demonstrate that MENAS achieves state-of-the-art performance.

arXiv:2203.04781

How many Observations are Enough? Knowledge Distillation for Trajectory Forecasting

Accurate prediction of future human positions is an essential task for modern video-surveillance systems. Current state-of-the-art models usually rely on a "history" of past tracked locations (e.g., 3 to 5 seconds) to predict a plausible sequence of future locations (e.g., up to the next 5 seconds). We feel that this common schema neglects critical traits of realistic applications: as the collection of input trajectories involves machine perception (i.e., detection and tracking), incorrect detection and fragmentation errors may accumulate in crowded scenes, leading to tracking drifts. On this account, the model would be fed with corrupted and noisy input data, thus fatally affecting its prediction performance. In this regard, we focus on delivering accurate predictions when only few input observations are used, thus potentially lowering the risks associated with automatic perception. To this end, we conceive a novel distillation strategy that allows a knowledge transfer from a teacher network to a student one, the latter fed with fewer observations (just two ones). We show that a properly defined teacher supervision allows a student network to perform comparably to state-of-the-art approaches that demand more observations. Besides, extensive experiments on common trajectory forecasting datasets highlight that our student network better generalizes to unseen scenarios.

arXiv:2203.03970

On Generalizing Beyond Domains in Cross-Domain Continual Learning

Humans have the ability to accumulate knowledge of new tasks in varying conditions, but deep neural networks often suffer from catastrophic forgetting of previously learned knowledge after learning a new task. Many recent methods focus on preventing catastrophic forgetting under the assumption of train and test data following similar distributions. In this work, we consider a more realistic scenario of continual learning under domain shifts where the model must generalize its inference to an unseen domain. To this end, we encourage learning semantically meaningful features by equipping the classifier with class similarity metrics as learning parameters which are obtained through Mahalanobis similarity computations. Learning of the backbone representation along with these extra parameters is done seamlessly in an end-to-end manner. In addition, we propose an approach based on the exponential moving average of the parameters for better knowledge distillation. We demonstrate that, to a great extent, existing continual learning algorithms fail to handle the forgetting issue under multiple distributions, while our proposed approach learns new tasks under domain shift with accuracy boosts up to 10% on challenging datasets such as DomainNet and OfficeHome.

arXiv:2203.03631공식 코드 · Papers with Code

Student Becomes Decathlon Master in Retinal Vessel Segmentation via Dual-teacher Multi-target Domain Adaptation

Unsupervised domain adaptation has been proposed recently to tackle the so-called domain shift between training data and test data with different distributions. However, most of them only focus on single-target domain adaptation and cannot be applied to the scenario with multiple target domains. In this paper, we propose RVms, a novel unsupervised multi-target domain adaptation approach to segment retinal vessels (RVs) from multimodal and multicenter retinal images. RVms mainly consists of a style augmentation and transfer (SAT) module and a dual-teacher knowledge distillation (DTKD) module. SAT augments and clusters images into source-similar domains and source-dissimilar domains via Bezier and Fourier transformations. DTKD utilizes the augmented and transformed data to train two teachers, one for source-similar domains and the other for source-dissimilar domains. Afterwards, knowledge distillation is performed to iteratively distill different domain knowledge from teachers to a generic student. The local relative intensity transformation is employed to characterize RVs in a domain invariant manner and promote the generalizability of teachers and student models. Moreover, we construct a new multimodal and multicenter vascular segmentation dataset from existing publicly-available datasets, which can be used to benchmark various domain adaptation and domain generalization methods. Through extensive experiments, RVms is found to be very close to the target-trained Oracle in terms of segmenting the RVs, largely outperforming other state-of-the-art methods.

arXiv:2203.03910공식 코드 · Papers with Code

Overcoming Catastrophic Forgetting beyond Continual Learning: Balanced Training for Neural Machine Translation

Neural networks tend to gradually forget the previously learned knowledge when learning multiple tasks sequentially from dynamic data distributions. This problem is called \textit{catastrophic forgetting}, which is a fundamental challenge in the continual learning of neural networks. In this work, we observe that catastrophic forgetting not only occurs in continual learning but also affects the traditional static training. Neural networks, especially neural machine translation models, suffer from catastrophic forgetting even if they learn from a static training set. To be specific, the final model pays imbalanced attention to training samples, where recently exposed samples attract more attention than earlier samples. The underlying cause is that training samples do not get balanced training in each model update, so we name this problem \textit{imbalanced training}. To alleviate this problem, we propose Complementary Online Knowledge Distillation (COKD), which uses dynamically updated teacher models trained on specific data orders to iteratively provide complementary knowledge to the student model. Experimental results on multiple machine translation tasks show that our method successfully alleviates the problem of imbalanced training and achieves substantial improvements over strong baseline systems.

arXiv:2203.03218공식 코드 · Papers with Code

Enhance Language Identification using Dual-mode Model with Knowledge Distillation

In this paper, we propose to employ a dual-mode framework on the x-vector self-attention (XSA-LID) model with knowledge distillation (KD) to enhance its language identification (LID) performance for both long and short utterances. The dual-mode XSA-LID model is trained by jointly optimizing both the full and short modes with their respective inputs being the full-length speech and its short clip extracted by a specific Boolean mask, and KD is applied to further boost the performance on short utterances. In addition, we investigate the impact of clip-wise linguistic variability and lexical integrity for LID by analyzing the variation of LID performance in terms of the lengths and positions of the mimicked speech clips. We evaluated our approach on the MLS14 data from the NIST 2017 LRE. With the 3~s random-location Boolean mask, our proposed method achieved 19.23%, 21.52% and 8.37% relative improvement in average cost compared with the XSA-LID model on 3s, 10s, and 30s speech, respectively.

arXiv:2203.02721공식 코드 · Papers with Code

Consistent Representation Learning for Continual Relation Extraction

Continual relation extraction (CRE) aims to continuously train a model on data with new relations while avoiding forgetting old ones. Some previous work has proved that storing a few typical samples of old relations and replaying them when learning new relations can effectively avoid forgetting. However, these memory-based methods tend to overfit the memory samples and perform poorly on imbalanced datasets. To solve these challenges, a consistent representation learning method is proposed, which maintains the stability of the relation embedding by adopting contrastive learning and knowledge distillation when replaying memory. Specifically, supervised contrastive learning based on a memory bank is first used to train each new task so that the model can effectively learn the relation representation. Then, contrastive replay is conducted of the samples in memory and makes the model retain the knowledge of historical relations through memory knowledge distillation to prevent the catastrophic forgetting of the old task. The proposed method can better learn consistent representations to alleviate forgetting effectively. Extensive experiments on FewRel and TACRED datasets show that our method significantly outperforms state-of-the-art baselines and yield strong robustness on the imbalanced dataset.

arXiv:2203.02651공식 코드 · Papers with Code

Ensemble Knowledge Guided Sub-network Search and Fine-tuning for Filter Pruning

Conventional NAS-based pruning algorithms aim to find the sub-network with the best validation performance. However, validation performance does not successfully represent test performance, i.e., potential performance. Also, although fine-tuning the pruned network to restore the performance drop is an inevitable process, few studies have handled this issue. This paper provides a novel Ensemble Knowledge Guidance (EKG) to solve both problems at once. First, we experimentally prove that the fluctuation of loss landscape can be an effective metric to evaluate the potential performance. In order to search a sub-network with the smoothest loss landscape at a low cost, we employ EKG as a search reward. EKG utilized for the following search iteration is composed of the ensemble knowledge of interim sub-networks, i.e., the by-products of the sub-network evaluation. Next, we reuse EKG to provide a gentle and informative guidance to the pruned network while fine-tuning the pruned network. Since EKG is implemented as a memory bank in both phases, it requires a negligible cost. For example, when pruning and training ResNet-50, just 315 GPU hours are required to remove around 45.04% of FLOPS without any performance degradation, which can operate even on a low-spec workstation. the implemented code is available at https://github.com/sseung0703/EKG.

arXiv:2203.02485공식 코드 · Papers with Code

Better Supervisory Signals by Observing Learning Paths

Better-supervised models might have better performance. In this paper, we first clarify what makes for good supervision for a classification problem, and then explain two existing label refining methods, label smoothing and knowledge distillation, in terms of our proposed criterion. To further answer why and how better supervision emerges, we observe the learning path, i.e., the trajectory of the model's predictions during training, for each training sample. We find that the model can spontaneously refine "bad" labels through a "zig-zag" learning path, which occurs on both toy and real datasets. Observing the learning path not only provides a new perspective for understanding knowledge distillation, overfitting, and learning dynamics, but also reveals that the supervisory signal of a teacher network can be very unstable near the best points in training on real tasks. Inspired by this, we propose a new knowledge distillation scheme, Filter-KD, which improves downstream classification performance in various settings.

arXiv:2203.00915

MIAShield: Defending Membership Inference Attacks via Preemptive Exclusion of Members

In membership inference attacks (MIAs), an adversary observes the predictions of a model to determine whether a sample is part of the model's training data. Existing MIA defenses conceal the presence of a target sample through strong regularization, knowledge distillation, confidence masking, or differential privacy. We propose MIAShield, a new MIA defense based on preemptive exclusion of member samples instead of masking the presence of a member. The key insight in MIAShield is weakening the strong membership signal that stems from the presence of a target sample by preemptively excluding it at prediction time without compromising model utility. To that end, we design and evaluate a suite of preemptive exclusion oracles leveraging model-confidence, exact or approximate sample signature, and learning-based exclusion of member data points. To be practical, MIAShield splits a training data into disjoint subsets and trains each subset to build an ensemble of models. The disjointedness of subsets ensures that a target sample belongs to only one subset, which isolates the sample to facilitate the preemptive exclusion goal. We evaluate MIAShield on three benchmark image classification datasets. We show that MIAShield effectively mitigates membership inference (near random guess) for a wide range of MIAs, achieves far better privacy-utility trade-off compared with state-of-the-art defenses, and remains resilient against an adaptive adversary.

arXiv:2203.00843공식 코드 · Papers with Code

X-Trans2Cap: Cross-Modal Knowledge Transfer using Transformer for 3D Dense Captioning

3D dense captioning aims to describe individual objects by natural language in 3D scenes, where 3D scenes are usually represented as RGB-D scans or point clouds. However, only exploiting single modal information, e.g., point cloud, previous approaches fail to produce faithful descriptions. Though aggregating 2D features into point clouds may be beneficial, it introduces an extra computational burden, especially in inference phases. In this study, we investigate a cross-modal knowledge transfer using Transformer for 3D dense captioning, X-Trans2Cap, to effectively boost the performance of single-modal 3D caption through knowledge distillation using a teacher-student framework. In practice, during the training phase, the teacher network exploits auxiliary 2D modality and guides the student network that only takes point clouds as input through the feature consistency constraints. Owing to the well-designed cross-modal feature fusion module and the feature alignment in the training phase, X-Trans2Cap acquires rich appearance information embedded in 2D images with ease. Thus, a more faithful caption can be generated only using point clouds during the inference. Qualitative and quantitative results confirm that X-Trans2Cap outperforms previous state-of-the-art by a large margin, i.e., about +21 and about +16 absolute CIDEr score on ScanRefer and Nr3D datasets, respectively.

arXiv:2203.00236

TRILLsson: Distilled Universal Paralinguistic Speech Representations

Recent advances in self-supervision have dramatically improved the quality of speech representations. However, deployment of state-of-the-art embedding models on devices has been restricted due to their limited public availability and large resource footprint. Our work addresses these issues by publicly releasing a collection of paralinguistic speech models that are small and near state-of-the-art performance. Our approach is based on knowledge distillation, and our models are distilled on public data only. We explore different architectures and thoroughly evaluate our models on the Non-Semantic Speech (NOSS) benchmark. Our largest distilled model is less than 15% the size of the original model (314MB vs 2.2GB), achieves over 96% the accuracy on 6 of 7 tasks, and is trained on 6.5% the data. The smallest model is 1% in size (22MB) and achieves over 90% the accuracy on 6 of 7 tasks. Our models outperform the open source Wav2Vec 2.0 model on 6 of 7 tasks, and our smallest model outperforms the open source Wav2Vec 2.0 on both emotion recognition tasks despite being 7% the size.

arXiv:2203.00585공식 코드 · Papers with Code

Self-Supervised Vision Transformers Learn Visual Concepts in Histopathology

Tissue phenotyping is a fundamental task in learning objective characterizations of histopathologic biomarkers within the tumor-immune microenvironment in cancer pathology. However, whole-slide imaging (WSI) is a complex computer vision in which: 1) WSIs have enormous image resolutions with precludes large-scale pixel-level efforts in data curation, and 2) diversity of morphological phenotypes results in inter- and intra-observer variability in tissue labeling. To address these limitations, current efforts have proposed using pretrained image encoders (transfer learning from ImageNet, self-supervised pretraining) in extracting morphological features from pathology, but have not been extensively validated. In this work, we conduct a search for good representations in pathology by training a variety of self-supervised models with validation on a variety of weakly-supervised and patch-level tasks. Our key finding is in discovering that Vision Transformers using DINO-based knowledge distillation are able to learn data-efficient and interpretable features in histology images wherein the different attention heads learn distinct morphological phenotypes. We make evaluation code and pretrained weights publicly-available at: https://github.com/Richarizardd/Self-Supervised-ViT-Path.

arXiv:2203.00600

Dual Embodied-Symbolic Concept Representations for Deep Learning

Motivated by recent findings from cognitive neural science, we advocate the use of a dual-level model for concept representations: the embodied level consists of concept-oriented feature representations, and the symbolic level consists of concept graphs. Embodied concept representations are modality specific and exist in the form of feature vectors in a feature space. Symbolic concept representations, on the other hand, are amodal and language specific, and exist in the form of word / knowledge-graph embeddings in a concept / knowledge space. The human conceptual system comprises both embodied representations and symbolic representations, which typically interact to drive conceptual processing. As such, we further advocate the use of dual embodied-symbolic concept representations for deep learning. To demonstrate their usage and value, we discuss two important use cases: embodied-symbolic knowledge distillation for few-shot class incremental learning, and embodied-symbolic fused representation for image-text matching. Dual embodied-symbolic concept representations are the foundation for deep learning and symbolic AI integration. We discuss two important examples of such integration: scene graph generation with knowledge graph bridging, and multimodal knowledge graphs.

arXiv:2202.13393공식 코드 · Papers with Code

TransKD: Transformer Knowledge Distillation for Efficient Semantic Segmentation

Semantic segmentation benchmarks in the realm of autonomous driving are dominated by large pre-trained transformers, yet their widespread adoption is impeded by substantial computational costs and prolonged training durations. To lift this constraint, we look at efficient semantic segmentation from a perspective of comprehensive knowledge distillation and aim to bridge the gap between multi-source knowledge extractions and transformer-specific patch embeddings. We put forward the Transformer-based Knowledge Distillation (TransKD) framework which learns compact student transformers by distilling both feature maps and patch embeddings of large teacher transformers, bypassing the long pre-training process and reducing the FLOPs by >85.0%. Specifically, we propose two fundamental modules to realize feature map distillation and patch embedding distillation, respectively: (1) Cross Selective Fusion (CSF) enables knowledge transfer between cross-stage features via channel attention and feature map distillation within hierarchical transformers; (2) Patch Embedding Alignment (PEA) performs dimensional transformation within the patchifying process to facilitate the patch embedding distillation. Furthermore, we introduce two optimization modules to enhance the patch embedding distillation from different perspectives: (1) Global-Local Context Mixer (GL-Mixer) extracts both global and local information of a representative embedding; (2) Embedding Assistant (EA) acts as an embedding method to seamlessly bridge teacher and student models with the teacher's number of channels. Experiments on Cityscapes, ACDC, NYUv2, and Pascal VOC2012 datasets show that TransKD outperforms state-of-the-art distillation frameworks and rivals the time-consuming pre-training method. The source code is publicly available at https://github.com/RuipingL/TransKD.

arXiv:2202.13123공식 코드 · Papers with Code

Content-Variant Reference Image Quality Assessment via Knowledge Distillation

Generally, humans are more skilled at perceiving differences between high-quality (HQ) and low-quality (LQ) images than directly judging the quality of a single LQ image. This situation also applies to image quality assessment (IQA). Although recent no-reference (NR-IQA) methods have made great progress to predict image quality free from the reference image, they still have the potential to achieve better performance since HQ image information is not fully exploited. In contrast, full-reference (FR-IQA) methods tend to provide more reliable quality evaluation, but its practicability is affected by the requirement for pixel-level aligned reference images. To address this, we firstly propose the content-variant reference method via knowledge distillation (CVRKD-IQA). Specifically, we use non-aligned reference (NAR) images to introduce various prior distributions of high-quality images. The comparisons of distribution differences between HQ and LQ images can help our model better assess the image quality. Further, the knowledge distillation transfers more HQ-LQ distribution difference information from the FR-teacher to the NAR-student and stabilizing CVRKD-IQA performance. Moreover, to fully mine the local-global combined information, while achieving faster inference speed, our model directly processes multiple image patches from the input with the MLP-mixer. Cross-dataset experiments verify that our model can outperform all NAR/NR-IQA SOTAs, even reach comparable performance with FR-IQA methods on some occasions. Since the content-variant and non-aligned reference HQ images are easy to obtain, our model can support more IQA applications with its relative robustness to content variations. Our code and more detailed elaborations of supplements are available: https://github.com/guanghaoyin/CVRKD-IQA.