PyTorchKR Papers

최신 논문

총 10,240편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2304.12314공식 코드 · Papers with Code

Distilling from Similar Tasks for Transfer Learning on a Budget

We address the challenge of getting efficient yet accurate recognition systems with limited labels. While recognition models improve with model size and amount of data, many specialized applications of computer vision have severe resource constraints both during training and inference. Transfer learning is an effective solution for training with few labels, however often at the expense of a computationally costly fine-tuning of large base models. We propose to mitigate this unpleasant trade-off between compute and accuracy via semi-supervised cross-domain distillation from a set of diverse source models. Initially, we show how to use task similarity metrics to select a single suitable source model to distill from, and that a good selection process is imperative for good downstream performance of a target model. We dub this approach DistillNearest. Though effective, DistillNearest assumes a single source model matches the target task, which is not always the case. To alleviate this, we propose a weighted multi-source distillation method to distill multiple source models trained on different domains weighted by their relevance for the target task into a single efficient model (named DistillWeighted). Our methods need no access to source data, and merely need features and pseudo-labels of the source models. When the goal is accurate recognition under computational constraints, both DistillNearest and DistillWeighted approaches outperform both transfer learning from strong ImageNet initializations as well as state-of-the-art semi-supervised techniques such as FixMatch. Averaged over 8 diverse target tasks our multi-source method outperforms the baselines by 5.6%-points and 4.5%-points, respectively.

arXiv:2304.11049

Using Mobile Data and Deep Models to Assess Auditory Verbal Hallucinations

Hallucination is an apparent perception in the absence of real external sensory stimuli. An auditory hallucination is a perception of hearing sounds that are not real. A common form of auditory hallucination is hearing voices in the absence of any speakers which is known as Auditory Verbal Hallucination (AVH). AVH is fragments of the mind's creation that mostly occur in people diagnosed with mental illnesses such as bipolar disorder and schizophrenia. Assessing the valence of hallucinated voices (i.e., how negative or positive voices are) can help measure the severity of a mental illness. We study N=435 individuals, who experience hearing voices, to assess auditory verbal hallucination. Participants report the valence of voices they hear four times a day for a month through ecological momentary assessments with questions that have four answering scales from ``not at all'' to ``extremely''. We collect these self-reports as the valence supervision of AVH events via a mobile application. Using the application, participants also record audio diaries to describe the content of hallucinated voices verbally. In addition, we passively collect mobile sensing data as contextual signals. We then experiment with how predictive these linguistic and contextual cues from the audio diary and mobile sensing data are of an auditory verbal hallucination event. Finally, using transfer learning and data fusion techniques, we train a neural net model that predicts the valance of AVH with a performance of 54\% top-1 and 72\% top-2 F1 score.

arXiv:2304.11004

Knowledge Distillation Under Ideal Joint Classifier Assumption

Knowledge distillation constitutes a potent methodology for condensing substantial neural networks into more compact and efficient counterparts. Within this context, softmax regression representation learning serves as a widely embraced approach, leveraging a pre-established teacher network to guide the learning process of a diminutive student network. Notably, despite the extensive inquiry into the efficacy of softmax regression representation learning, the intricate underpinnings governing the knowledge transfer mechanism remain inadequately elucidated. This study introduces the 'Ideal Joint Classifier Knowledge Distillation' (IJCKD) framework, an overarching paradigm that not only furnishes a lucid and exhaustive comprehension of prevailing knowledge distillation techniques but also establishes a theoretical underpinning for prospective investigations. Employing mathematical methodologies derived from domain adaptation theory, this investigation conducts a comprehensive examination of the error boundary of the student network contingent upon the teacher network. Consequently, our framework facilitates efficient knowledge transference between teacher and student networks, thereby accommodating a diverse spectrum of applications.

arXiv:2304.10767공식 코드 · Papers with Code

How good are variational autoencoders at transfer learning?

Variational autoencoders (VAEs) are used for transfer learning across various research domains such as music generation or medical image analysis. However, there is no principled way to assess before transfer which components to retrain or whether transfer learning is likely to help on a target task. We propose to explore this question through the lens of representational similarity. Specifically, using Centred Kernel Alignment (CKA) to evaluate the similarity of VAEs trained on different datasets, we show that encoders' representations are generic but decoders' specific. Based on these insights, we discuss the implications for selecting which components of a VAE to retrain and propose a method to visually assess whether transfer learning is likely to help on classification tasks.

arXiv:2304.10880

Med-Tuning: A New Parameter-Efficient Tuning Framework for Medical Volumetric Segmentation

The "pre-training then fine-tuning (FT)" paradigm is widely adopted to boost the model performance of deep learning-based methods for medical volumetric segmentation. However, conventional full FT incurs high computational and memory costs. Thus, it is of increasing importance to fine-tune pre-trained models for medical volumetric segmentation tasks in a both effective and parameter-efficient manner. In this paper, we introduce a new framework named Med-Tuning to realize parameter-efficient tuning (PET) for medical volumetric segmentation task and an efficient plug-and-play module named Med-Adapter for task-specific feature extraction. With a small number of tuned parameters, our framework enhances the 2D baselines's precision on segmentation tasks, which are pre-trained on natural images. Extensive experiments on three benchmark datasets (CT and MRI modalities) show that our method achieves better results than previous PET methods on volumetric segmentation tasks. Compared to full FT, Med-Tuning reduces the fine-tuned model parameters by up to 4x, with even better segmentation performance. Our project webpage is at \url{https://rubics-xuan.github.io/Med-Tuning/}.

arXiv:2304.10597공식 코드 · Papers with Code

Text2Seg: Remote Sensing Image Semantic Segmentation via Text-Guided Visual Foundation Models

Remote sensing imagery has attracted significant attention in recent years due to its instrumental role in global environmental monitoring, land usage monitoring, and more. As image databases grow each year, performing automatic segmentation with deep learning models has gradually become the standard approach for processing the data. Despite the improved performance of current models, certain limitations remain unresolved. Firstly, training deep learning models for segmentation requires per-pixel annotations. Given the large size of datasets, only a small portion is fully annotated and ready for training. Additionally, the high intra-dataset variance in remote sensing data limits the transfer learning ability of such models. Although recently proposed generic segmentation models like SAM have shown promising results in zero-shot instance-level segmentation, adapting them to semantic segmentation is a non-trivial task. To tackle these challenges, we propose a novel method named Text2Seg for remote sensing semantic segmentation. Text2Seg overcomes the dependency on extensive annotations by employing an automatic prompt generation process using different visual foundation models (VFMs), which are trained to understand semantic information in various ways. This approach not only reduces the need for fully annotated datasets but also enhances the model's ability to generalize across diverse datasets. Evaluations on four widely adopted remote sensing datasets demonstrate that Text2Seg significantly improves zero-shot prediction performance compared to the vanilla SAM model, with relative improvements ranging from 31% to 225%. Our code is available at https://github.com/Douglas2Code/Text2Seg.

arXiv:2304.10550

Deep transfer learning for intrusion detection in industrial control networks: A comprehensive review

Globally, the external internet is increasingly being connected to industrial control systems. As a result, there is an immediate need to protect these networks from a variety of threats. The key infrastructure of industrial activity can be protected from harm using an intrusion detection system (IDS), a preventive mechanism that seeks to recognize new kinds of dangerous threats and hostile activities. This review examines the most recent artificial-intelligence techniques that are used to create IDSs in many kinds of industrial control networks, with a particular emphasis on IDS-based deep transfer learning (DTL). DTL can be seen as a type of information-fusion approach that merges and/or adapts knowledge from multiple domains to enhance the performance of a target task, particularly when labeled data in the target domain is scarce. Publications issued after 2015 were considered. These selected publications were divided into three categories: DTL-only and IDS-only works are examined in the introduction and background section, and DTL-based IDS papers are considered in the core section of this review. By reading this review paper, researchers will be able to gain a better grasp of the current state of DTL approaches used in IDSs in many different types of network. Other useful information, such as the datasets used, the type of DTL employed, the pre-trained network, IDS techniques, the evaluation metrics including accuracy/F-score and false-alarm rate, and the improvements gained, are also covered. The algorithms and methods used in several studies are presented, and the principles of DTL-based IDS subcategories are presented to the reader and illustrated deeply and clearly

arXiv:2304.10739공식 코드 · Papers with Code

KitchenScale: Learning to predict ingredient quantities from recipe contexts

Determining proper quantities for ingredients is an essential part of cooking practice from the perspective of enriching tastiness and promoting healthiness. We introduce KitchenScale, a fine-tuned Pre-trained Language Model (PLM) that predicts a target ingredient's quantity and measurement unit given its recipe context. To effectively train our KitchenScale model, we formulate an ingredient quantity prediction task that consists of three sub-tasks which are ingredient measurement type classification, unit classification, and quantity regression task. Furthermore, we utilized transfer learning of cooking knowledge from recipe texts to PLMs. We adopted the Discrete Latent Exponent (DExp) method to cope with high variance of numerical scales in recipe corpora. Experiments with our newly constructed dataset and recommendation examples demonstrate KitchenScale's understanding of various recipe contexts and generalizability in predicting ingredient quantities. We implemented a web application for KitchenScale to demonstrate its functionality in recommending ingredient quantities expressed in numerals (e.g., 2) with units (e.g., ounce).

arXiv:2304.10464공식 코드 · Papers with Code

Learning to Plan with Natural Language

Large Language Models (LLMs) have shown remarkable performance in various basic natural language tasks. For completing the complex task, we still need a plan for the task to guide LLMs to generate the specific solutions step by step. LLMs can directly generate task plans, but these plans may still contain factual errors or are incomplete. A high-quality task plan contains correct step-by-step solutions for solving all situations and behavioral instructions for avoiding mistakes. To obtain it, we propose the Learning to Plan method, which involves two phases: (1) In the first learning task plan phase, it iteratively updates the task plan with new step-by-step solutions and behavioral instructions, which are obtained by prompting LLMs to derive from training error feedback. (2) In the subsequent test phase, the LLM uses the learned task plan to guide the inference of LLM on the test set. We demonstrate the effectiveness of our method on the five different reasoning type tasks (8 datasets). Further, our analysis experiment shows that the task plan learned by one LLM can directly guide another LLM to improve its performance, which reveals a new transfer learning paradigm. We release the code at \url{https://github.com/Eureka6174/LearnNLPlan}

arXiv:2304.10295

Decouple Non-parametric Knowledge Distillation For End-to-end Speech Translation

Existing techniques often attempt to make knowledge transfer from a powerful machine translation (MT) to speech translation (ST) model with some elaborate techniques, which often requires transcription as extra input during training. However, transcriptions are not always available, and how to improve the ST model performance without transcription, i.e., data efficiency, has rarely been studied in the literature. In this paper, we propose Decoupled Non-parametric Knowledge Distillation (DNKD) from data perspective to improve the data efficiency. Our method follows the knowledge distillation paradigm. However, instead of obtaining the teacher distribution from a sophisticated MT model, we construct it from a non-parametric datastore via k-Nearest-Neighbor (kNN) retrieval, which removes the dependence on transcription and MT model. Then we decouple the classic knowledge distillation loss into target and non-target distillation to enhance the effect of the knowledge among non-target logits, which is the prominent "dark knowledge". Experiments on MuST-C corpus show that, the proposed method can achieve consistent improvement over the strong baseline without requiring any transcription.

arXiv:2304.10036

Visual DNA: Representing and Comparing Images using Distributions of Neuron Activations

Selecting appropriate datasets is critical in modern computer vision. However, no general-purpose tools exist to evaluate the extent to which two datasets differ. For this, we propose representing images - and by extension datasets - using Distributions of Neuron Activations (DNAs). DNAs fit distributions, such as histograms or Gaussians, to activations of neurons in a pre-trained feature extractor through which we pass the image(s) to represent. This extractor is frozen for all datasets, and we rely on its generally expressive power in feature space. By comparing two DNAs, we can evaluate the extent to which two datasets differ with granular control over the comparison attributes of interest, providing the ability to customise the way distances are measured to suit the requirements of the task at hand. Furthermore, DNAs are compact, representing datasets of any size with less than 15 megabytes. We demonstrate the value of DNAs by evaluating their applicability on several tasks, including conditional dataset comparison, synthetic image evaluation, and transfer learning, and across diverse datasets, ranging from synthetic cat images to celebrity faces and urban driving scenes.

arXiv:2304.09976공식 코드 · Papers with Code

Analyzing the Domain Shift Immunity of Deep Homography Estimation

Homography estimation serves as a fundamental technique for image alignment in a wide array of applications. The advent of convolutional neural networks has introduced learning-based methodologies that have exhibited remarkable efficacy in this realm. Yet, the generalizability of these approaches across distinct domains remains underexplored. Unlike other conventional tasks, CNN-driven homography estimation models show a distinctive immunity to domain shifts, enabling seamless deployment from one dataset to another without the necessity of transfer learning. This study explores the resilience of a variety of deep homography estimation models to domain shifts, revealing that the network architecture itself is not a contributing factor to this remarkable adaptability. By closely examining the models' focal regions and subjecting input images to a variety of modifications, we confirm that the models heavily rely on local textures such as edges and corner points for homography estimation. Moreover, our analysis underscores that the domain shift immunity itself is intricately tied to the utilization of these local textures.

arXiv:2304.09874공식 코드 · Papers with Code

Domain Adaptable Self-supervised Representation Learning on Remote Sensing Satellite Imagery

This work presents a novel domain adaption paradigm for studying contrastive self-supervised representation learning and knowledge transfer using remote sensing satellite data. Major state-of-the-art remote sensing visual domain efforts primarily focus on fully supervised learning approaches that rely entirely on human annotations. On the other hand, human annotations in remote sensing satellite imagery are always subject to limited quantity due to high costs and domain expertise, making transfer learning a viable alternative. The proposed approach investigates the knowledge transfer of selfsupervised representations across the distinct source and target data distributions in depth in the remote sensing data domain. In this arrangement, self-supervised contrastive learning-based pretraining is performed on the source dataset, and downstream tasks are performed on the target datasets in a round-robin fashion. Experiments are conducted on three publicly available datasets, UC Merced Landuse (UCMD), SIRI-WHU, and MLRSNet, for different downstream classification tasks versus label efficiency. In self-supervised knowledge transfer, the proposed approach achieves state-of-the-art performance with label efficiency labels and outperforms a fully supervised setting. A more in-depth qualitative examination reveals consistent evidence for explainable representation learning. The source code and trained models are published on GitHub.

arXiv:2304.10179공식 코드 · Papers with Code

SCoDA: Domain Adaptive Shape Completion for Real Scans

3D shape completion from point clouds is a challenging task, especially from scans of real-world objects. Considering the paucity of 3D shape ground truths for real scans, existing works mainly focus on benchmarking this task on synthetic data, e.g. 3D computer-aided design models. However, the domain gap between synthetic and real data limits the generalizability of these methods. Thus, we propose a new task, SCoDA, for the domain adaptation of real scan shape completion from synthetic data. A new dataset, ScanSalon, is contributed with a bunch of elaborate 3D models created by skillful artists according to scans. To address this new task, we propose a novel cross-domain feature fusion method for knowledge transfer and a novel volume-consistent self-training framework for robust learning from real data. Extensive experiments prove our method is effective to bring an improvement of 6%~7% mIoU.

arXiv:2304.10309

Improving Speech Translation by Cross-Modal Multi-Grained Contrastive Learning

The end-to-end speech translation (E2E-ST) model has gradually become a mainstream paradigm due to its low latency and less error propagation. However, it is non-trivial to train such a model well due to the task complexity and data scarcity. The speech-and-text modality differences result in the E2E-ST model performance usually inferior to the corresponding machine translation (MT) model. Based on the above observation, existing methods often use sharingmechanisms to carry out implicit knowledge transfer by imposing various constraints. However, the final model often performs worse on the MT task than the MT model trained alone, which means that the knowledge transfer ability of this method is also limited. To deal with these problems, we propose the FCCL (Fine- and Coarse- Granularity Contrastive Learning) approach for E2E-ST, which makes explicit knowledge transfer through cross-modal multi-grained contrastive learning. A key ingredient of our approach is applying contrastive learning at both sentence- and frame-level to give the comprehensive guide for extracting speech representations containing rich semantic information.In addition, we adopt a simple whitening method to alleviate the representation degeneration in the MT model, which adversely affects contrast learning. Experiments on the MuST-C benchmark show that our proposed approach significantly outperforms the state-of-the-art E2E-ST baselines on all eight language pairs. Further analysis indicates that FCCL can free up its capacity from learning grammatical structure information and force more layers to learn semantic information.

arXiv:2304.08766

Cashew dataset generation using augmentation and RaLSGAN and a transfer learning based tinyML approach towards disease detection

Cashew is one of the most extensively consumed nuts in the world, and it is also known as a cash crop. A tree may generate a substantial yield in a few months and has a lifetime of around 70 to 80 years. Yet, in addition to the benefits, there are certain constraints to its cultivation. With the exception of parasites and algae, anthracnose is the most common disease affecting trees. When it comes to cashew, the dense structure of the tree makes it difficult to diagnose the disease with ease compared to short crops. Hence, we present a dataset that exclusively consists of healthy and diseased cashew leaves and fruits. The dataset is authenticated by adding RGB color transformation to highlight diseased regions, photometric and geometric augmentations, and RaLSGAN to enlarge the initial collection of images and boost performance in real-time situations when working with a constrained dataset. Further, transfer learning is used to test the classification efficiency of the dataset using algorithms such as MobileNet and Inception. TensorFlow lite is utilized to develop these algorithms for disease diagnosis utilizing drones in real-time. Several post-training optimization strategies are utilized, and their memory size is compared. They have proven their effectiveness by delivering high accuracy (up to 99%) and a decrease in memory and latency, making them ideal for use in applications with limited resources.

arXiv:2304.09715

UniCal: a Single-Branch Transformer-Based Model for Camera-to-LiDAR Calibration and Validation

We introduce a novel architecture, UniCal, for Camera-to-LiDAR (C2L) extrinsic calibration which leverages self-attention mechanisms through a Transformer-based backbone network to infer the 6-degree of freedom (DoF) relative transformation between the sensors. Unlike previous methods, UniCal performs an early fusion of the input camera and LiDAR data by aggregating camera image channels and LiDAR mappings into a multi-channel unified representation before extracting their features jointly with a single-branch architecture. This single-branch architecture makes UniCal lightweight, which is desirable in applications with restrained resources such as autonomous driving. Through experiments, we show that UniCal achieves state-of-the-art results compared to existing methods. We also show that through transfer learning, weights learned on the calibration task can be applied to a calibration validation task without re-training the backbone.

arXiv:2304.09067공식 코드 · Papers with Code

Performance of GAN-based augmentation for deep learning COVID-19 image classification

The biggest challenge in the application of deep learning to the medical domain is the availability of training data. Data augmentation is a typical methodology used in machine learning when confronted with a limited data set. In a classical approach image transformations i.e. rotations, cropping and brightness changes are used. In this work, a StyleGAN2-ADA model of Generative Adversarial Networks is trained on the limited COVID-19 chest X-ray image set. After assessing the quality of generated images they are used to increase the training data set improving its balance between classes. We consider the multi-class classification problem of chest X-ray images including the COVID-19 positive class that hasn't been yet thoroughly explored in the literature. Results of transfer learning-based classification of COVID-19 chest X-ray images are presented. The performance of several deep convolutional neural network models is compared. The impact on the detection performance of classical image augmentations i.e. rotations, cropping, and brightness changes are studied. Furthermore, classical image augmentation is compared with GAN-based augmentation. The most accurate model is an EfficientNet-B0 with an accuracy of 90.2 percent, trained on a dataset with a simple class balancing. The GAN augmentation approach is found to be subpar to classical methods for the considered dataset.

arXiv:2304.08600공식 코드 · Papers with Code

RS2G: Data-Driven Scene-Graph Extraction and Embedding for Robust Autonomous Perception and Scenario Understanding

Effectively capturing intricate interactions among road users is of critical importance to achieving safe navigation for autonomous vehicles. While graph learning (GL) has emerged as a promising approach to tackle this challenge, existing GL models rely on predefined domain-specific graph extraction rules that often fail in real-world drastically changing scenarios. Additionally, these graph extraction rules severely impede the capability of existing GL methods to generalize knowledge across domains. To address this issue, we propose RoadScene2Graph (RS2G), an innovative autonomous scenario understanding framework with a novel data-driven graph extraction and modeling approach that dynamically captures the diverse relations among road users. Our evaluations demonstrate that on average RS2G outperforms the state-of-the-art (SOTA) rule-based graph extraction method by 4.47% and the SOTA deep learning model by 22.19% in subjective risk assessment. More importantly, RS2G delivers notably better performance in transferring knowledge gained from simulation environments to unseen real-world scenarios.

arXiv:2304.08799

Self-Supervised 3D Action Representation Learning with Skeleton Cloud Colorization

3D Skeleton-based human action recognition has attracted increasing attention in recent years. Most of the existing work focuses on supervised learning which requires a large number of labeled action sequences that are often expensive and time-consuming to annotate. In this paper, we address self-supervised 3D action representation learning for skeleton-based action recognition. We investigate self-supervised representation learning and design a novel skeleton cloud colorization technique that is capable of learning spatial and temporal skeleton representations from unlabeled skeleton sequence data. We represent a skeleton action sequence as a 3D skeleton cloud and colorize each point in the cloud according to its temporal and spatial orders in the original (unannotated) skeleton sequence. Leveraging the colorized skeleton point cloud, we design an auto-encoder framework that can learn spatial-temporal features from the artificial color labels of skeleton joints effectively. Specifically, we design a two-steam pretraining network that leverages fine-grained and coarse-grained colorization to learn multi-scale spatial-temporal features. In addition, we design a Masked Skeleton Cloud Repainting task that can pretrain the designed auto-encoder framework to learn informative representations. We evaluate our skeleton cloud colorization approach with linear classifiers trained under different configurations, including unsupervised, semi-supervised, fully-supervised, and transfer learning settings. Extensive experiments on NTU RGB+D, NTU RGB+D 120, PKU-MMD, NW-UCLA, and UWA3D datasets show that the proposed method outperforms existing unsupervised and semi-supervised 3D action recognition methods by large margins and achieves competitive performance in supervised 3D action recognition as well.