Stochastic sampling algorithms for state estimation of jump Markov linear systems

Learning representational invariances for data-efficient action recognition

Much research on object detection focuses on building better model architectures and detection algorithms. Changing the model architecture, however, comes at the cost of adding more complexity to inference, making models slower. Data augmentation, on the other hand, doesn’t add any inference complexity, but is insufficiently studied in object detection for two reasons. First it is more difficult to design plausible augmentation strategies for object detection than for classification, because one must handle the complexity of bounding boxes if geometric transformations are applied. Secondly, data augmentation attracts less research attention perhaps because it is believed to add less value and to transfer poorly compared to advances in network architectures.This paper serves two main purposes. First, we propose to use AutoAugment [3] to design better data augmentation strategies for object detection because it can address the difficulty of designing them. Second, we use the method to assess the value of data augmentation in object detection and compare it against the value of architectures. Our investigation into data augmentation for object detection identifies two surprising results. First, by changing the data augmentation strategy to our method, AutoAugment for detection, we can improve RetinaNet with a ResNet-50 backbone from 36.7 to 39.0 mAP on COCO, a difference of +2.3 mAP. This gain exceeds the gain achieved by switching the backbone from ResNet-50 to ResNet-101 (+2.1 mAP), which incurs additional training and inference costs. The second surprising finding is that our strategies found on the COCO dataset transfer well to the PASCAL dataset to improve accuracy by +2.7 mAP. These results together with our systematic studies of data augmentation call into question previous assumptions about the role and transferability of architectures versus data augmentation. In particular, changing the augmentation may lead to performance gains that are equally transferable as changing the underlying architecture.

OBJECTIVE. Deep learning applications in radiology often suffer from overfitting, limiting generalization to external centers. The objective of this study was to develop a high-quality prostate segmentation model capable of maintaining a high degree of performance across multiple independent datasets using transfer learning and data augmentation. MATERIALS AND METHODS. A retrospective cohort of 648 patients who underwent prostate MRI between February 2015 and November 2018 at a single center was used for training and validation. A deep learning approach combining 2D and 3D architecture was used for training, which incorporated transfer learning. A data augmentation strategy was used that was specific to the deformations, intensity, and alterations in image quality seen on radiology images. Five independent datasets, four of which were from outside centers, were used for testing, which was conducted with and without fine-tuning of the original model. The Dice similarity coefficient was used to evaluate model performance. RESULTS. When prostate segmentation models utilizing transfer learning were applied to the internal validation cohort, the mean Dice similarity coefficient was 93.1 for whole prostate and 89.0 for transition zone segmentations. When the models were applied to multiple test set cohorts, the improvement in performance achieved using data augmentation alone was 2.2% for the whole prostate models and 3.0% for the transition zone segmentation models. However, the best test-set results were obtained with models fine-tuned on test center data with mean Dice similarity coefficients of 91.5 for whole prostate segmentation and 89.7 for transition zone segmentation. CONCLUSION. Transfer learning allowed for the development of a high-performing prostate segmentation model, and data augmentation and fine-tuning approaches improved performance of a prostate segmentation model when applied to datasets from external centers.

Data augmentation (DA) turns seemingly intractable computational problems into simple ones by augmenting latent missing data. In addition to computational simplicity, it is now well-established that DA equipped with a deterministic transformation can improve the convergence speed of iterative algorithms such as an EM algorithm or Gibbs sampler. In this article, we outline a framework for the transformation-based DA, which we call data transforming augmentation (DTA), allowing augmented data to be a deterministic function of latent and observed data, and unknown parameters. Under this framework, we investigate a novel DTA scheme that turns heteroscedastic models into homoscedastic ones to take advantage of simpler computations typically available in homoscedastic cases. Applying this DTA scheme to fitting linear mixed models, we demonstrate simpler computations and faster convergence rates of resulting iterative algorithms, compared with those under a non-transformation-based DA scheme. We also fit a Beta-Binomial model using the proposed DTA scheme, which enables sampling approximate marginal posterior distributions that are available only under homoscedasticity. Supplementary materials are available online.

EEG signals are the modality that is widely used to recognize human emotions. However, the limited data on EEG signals remains challenging because of the small recording participants, the need for an expert to interpret EEG signals, and the expensive cost of tools to record EEG signals. This research proposed the data augmentation schemes on the EEG datasets to overcome the limited available data problem. Augmenting the data will help the generalizability of the emotion recognition model. The EEG signals on the DEAP and SEED datasets are transformed into image samples using a recurrence plot and spectrogram. Then, the artificial recurrence plot and the artificial spectrogram samples are generated using Pix2pix. This research used these artificial samples to conduct the data augmentation process. LeNet5, ResNet50, MobileNet, and DenseNet121 are used to conduct the classification. The best four data augmentation schemes are as follows: Appending 20,000 artificial recurrence plot samples to DEAP and SEED training datasets, appending 20,000 artificial spectrogram samples to the DEAP training dataset, and appending 15,000 artificial spectrogram samples to the SEED training dataset. The kappa coefficient for each classification model based on the best data augmentation schemes is computed. It is found that among the compared classifiers, LeNet5 achieved the best accuracy in both SEED (98.58%) and DEAP (86.12%) datasets when spectrogram was used. Therefore, LeNet5 trained on the spectrogram samples is a reliable and robust classification model. This finding implies that the use of spectrogram is more promising than the recurrence plot in human emotion recognition.

With the fast-developing domain of cyber-physical systems (CPS), constructing the CPS with high-quality services becomes an imperative task. As one of the effective solutions for information overload in CPS construction, quality of service (QoS)-aware service recommendation has drawn much attention in academia and industry. However, the lack of most QoS values limits the recommendation performance and it is time-consuming for users to get the QoS values by invoking all the services. Therefore, a powerful prediction model is required to predict the unobserved QoS values. Considering the fact that most existing QoS prediction models are unable to effectively address the data-sparsity problem, a novel two-stage framework called AgQ is proposed for QoS prediction. Specifically, a data augmentation strategy is designed in the first stage to enlarge the training set by drawing additional virtual instances. In the second stage, a prediction model is applied that considers both virtual and factual instances during the training procedure. We conduct extensive experiments on the WSDream dataset to demonstrate the effectiveness of the our QoS prediction framework and verify that the data augmentation strategy can indeed alleviate the data-sparsity problem. In terms of mean absolute error, taking the Multilayer Perceptron model as an example, the maximum improvement achieves 5% under 5% sparsity.

Segmentation of tumor regions in H &E-stained slides is an important task for a pathologist while diagnosing different types of cancer, including oral squamous cell carcinoma (OSCC). Histological image segmentation is often constrained by the availability of labeled training data since labeling histological images is a highly skilled, complex, and time-consuming task. Thus, data augmentation strategies become essential to train convolutional neural networks models to overcome the overfitting problem when only a few training samples are available. This paper proposes a new data augmentation strategy, named Random Composition Augmentation (RCAug), to train fully convolutional networks (FCN) to segment OSCC tumor regions in H &E-stained histological images. Given the input image and their corresponding label, a pipeline with a random composition of geometric, distortion, color transfer, and generative image transformations is executed on the fly. Experimental evaluations were performed using an FCN-based method to segment OSCC regions through a set of different data augmentation transformations. By using RCAug, we improved the FCN-based segmentation method from 0.51 to 0.81 of intersection-over-union (IOU) in a whole slide image dataset and from 0.65 to 0.69 of IOU in a tissue microarray images dataset.

Decades of research have been dedicated to overcoming the obstacles inherent in synthetic aperture radar (SAR) automatic target recognition (ATR). The rise of deep learning technologies has brought a wave of new possibilities, demonstrating significant progress in the field. However, challenges like the susceptibility of SAR images to noise, the requirement for large-scale training datasets, and the often protracted duration of model training still persist. This paper introduces a novel data augmentation strategy to address these issues. Our method involves the intentional addition and subsequent removal of speckle noise to artificially enlarge the scope of training data through noise perturbation. Furthermore, we propose a modified network architecture named weighted ResNet, which incorporates residual strain controls for enhanced performance. This network is designed to be computationally efficient and to minimize the amount of training data required. Through rigorous experimental analysis, our research confirms that the proposed data augmentation method, when used in conjunction with the weighted ResNet model, significantly reduces the time needed for training. It also improves the SAR ATR capabilities. Compared to existing models and methods tested, the combination of our data augmentation scheme and the weighted ResNet framework achieves higher computational efficiency and better recognition accuracy in SAR ATR applications. This suggests that our approach could be a valuable advancement in the field of SAR image analysis.

Single-Domain Generalized Object Detection (S-DGOD) aims to train on a single source domain for robust performance across a variety of unseen target domains by taking advantage of an object detector. Existing S-DGOD approaches often rely on data augmentation strategies, including a composition of visual transformations, to enhance the detector's generalization ability. However, the absence of real-world prior knowledge hinders data augmentation from contributing to the diversity of training data distributions. To address this issue, we propose PhysAug, a novel physical model-based non-ideal imaging condition data augmentation method, to enhance the adaptability of the S-DGOD tasks. Drawing upon the principles of atmospheric optics, we develop a universal perturbation model that serves as the foundation for our proposed PhysAug. Given that visual perturbations typically arise from the interaction of light with atmospheric particles, the image frequency spectrum is harnessed to simulate real-world variations during training. This approach fosters the detector to learn domain-invariant representations, thereby enhancing its ability to generalize across various settings. Without altering the network architecture or loss function, our approach significantly outperforms the state-of-the-art across various S-DGOD datasets. In particular, it achieves a substantial improvement of 7.3% and 7.2% over the baseline on DWD and Cityscape-C, highlighting its enhanced generalizability in real-world settings.

The extensive utilization of drones has led to numerous scenarios that encompass both advantageous and perilous outcomes. By using deep learning techniques, this study aimed to reduce the dangerous effects of drone use through early detection of drones. The purpose of this study is the evaluation of deep learning approaches such as pre-trained YOLOv8 drone detection for security issues. This study focuses on the YOLOv8 model to achieve optimal performance in object detection tasks using a publicly available dataset collected by Mehdi Özel for a UAV competition that is sourced from GitHub. These images are labeled using Roboflow, and the model is trained on Google Colab. YOLOv8, known for its advanced architecture, was selected due to its suitability for real-time detection applications and its ability to process complex visual data. Hyperparameter tuning and data augmentation techniques were applied to maximize the performance of the model. Basic hyperparameters such as learning rate, batch size, and optimization settings were optimized through iterative experiments to provide the best performance. In addition to hyperparameter tuning, various data augmentation strategies were used to increase the robustness and generalization ability of the model. Techniques such as rotation, scaling, flipping, and color adjustments were applied to the dataset to simulate different conditions and variations. Among the augmentation techniques applied to the specific dataset in this study, rotation was found to deliver the highest performance. Blurring and cropping methods were observed to follow closely behind. The combination of optimized hyperparameters and strategic data augmentation allowed YOLOv8 to achieve high detection accuracy and reliable performance on the publicly available dataset. This method demonstrates the effectiveness of YOLOv8 in real-world scenarios, while also highlighting the importance of hyperparameter tuning and data augmentation in increasing model capabilities. To enhance model performance, dataset augmentation techniques including rotation and blurring are implemented. Following these steps, a significant precision value of 0.946, a notable recall value of 0.9605, and a considerable precision–recall curve value of 0.978 are achieved, surpassing many popular models such as Mask CNN, CNN, and YOLOv5.

Regularization graph convolutional networks with data augmentation

As an essential part of the telemedicine system, teleradiology not only realizes mutual recognition of image diagnosis among medical institutions, but also ensures the quality of medical image diagnosis. In this paper, an edge computing (EC) driven 5G teleradiology framework is proposed to segment glioma accurately. In our framework, through the high-speed transmission and sharing mechanism of medical image data under 5G condition, the high-definition magnetic resonance imaging (MRI) images data of glioma patients are delivered to the edge server for data augmentation and training. Three typical data augmentation methods, i.e., traditional geometric transformation, DCGAN, and CycleGAN, are used to study the impact of the proportions of synthetic images generated by the above data augmentation methods to original MRI images on the performance of glioma segmentation. It is observed that training with different proportions leads to different segmentation results. Based on this observation, an efficient data augmentation strategy is established, which exploits the optimal proportions of synthetic and original data to yield the best glioma segmentation results in terms of sensitivity, specificity, accuracy, and Dice score. Experiments on the BraTS dataset show that the proposed data augmentation strategy improves the average Dice score of Dense U-net compared to the conventional data augmentation method. With the aid of our EC driven 5G teleradiology framework, accurate glioma segmentation results can be acquired rapidly at the terminal equipment to facilitate the online diagnosis.

This study proposes a standardized framework for early prediction of gastrointestinal disease diseases leveraging advanced image pre-processing, data augmentation, and an attention-based fused Convolutional Neural Network (CNN). The framework initiates meticulous image pre-processing, encompassing cropping, standardization and resizing methodologies to standardize image inputs and accentuate pertinent features. Following pre-processing, data augmentation strategies are implemented to diversify the training dataset, enhancing model robustness and generalization. By artificially generating diverse variations of the original images, data augmentation helps expand the training dataset. Techniques such as rotation, translation, scaling, and flipping are applied to augment the dataset, enabling the CNN algorithm to generalize better and capture a broader range of disease patterns. The key component of the framework is an attention-based fused CNN architecture designed to capture spatial and channel-wise dependencies within gastrointestinal images effectively. The attention mechanism allows the network to concentrate on informative regions while suppressing irrelevant noise, enhancing feature representation and classification performance. The experimental outcomes proved that the developed framework attains superior performance in the early prediction of gastrointestinal diseases with a high accuracy of 98.9%. The combination of image pre-processing, data augmentation and attention-based fused CNN contributes to improved accuracy and robustness.

Deep learning is now widely used as an efficient tool for medical image classification and segmentation. However, conventional machine learning techniques are still more accurate than deep learning when only a small dataset is available. In this study, we present a general data augmentation strategy using Perlin noise, applying it to pixel-by-pixel image classification and quantification of various kinds of image patterns of diffuse interstitial lung disease (DILD). Using retrospectively obtained high-resolution computed tomography (HRCT) images from 106 patients, 100 regions-of-interest (ROIs) for each of six classes of image patterns (normal, ground-glass opacity, reticular opacity, honeycombing, emphysema, and consolidation) were selected for deep learning classification by experienced thoracic radiologists. For extra-validation, the deep learning quantification of the six classification patterns was evaluated for 92 HRCT whole lung images for which hand-labeled segmentation masks created by two experienced radiologists were available. FusionNet, a convolutional neural network (CNN), was used for training, test, and extra-validation on classifications of DILD image patterns. The accuracy of FusionNet with data augmentation using Perlin noise (89.5%, 49.8%, and 55.0% for ROI-based classification and whole lung quantifications by two radiologists, respectively) was significantly higher than that with conventional data augmentation (82.1%, 45.7%, and 49.9%, respectively). This data augmentation strategy using Perlin noise could be widely applied to deep learning studies for image classification and segmentation, especially in cases with relatively small datasets.

Enhanced Indonesian Ethnic Speaker Recognition using Data Augmentation Deep Neural Network