Search

The emergence of deep learning has considerably advanced the state-of-the-art in cardiac magnetic resonance (CMR) segmentation. Many techniques have been proposed over the last few years, bringing the accuracy of automated segmentation close to human performance. However, these models have been all too often trained and validated using cardiac imaging samples from single clinical centres or homogeneous imaging protocols. This has prevented the development and validation of models that are generalizable across different clinical centres, imaging conditions or scanner vendors. To promote further research and scientific benchmarking in the field of generalizable deep learning for cardiac segmentation, this paper presents the results of the Multi-Centre, Multi-Vendor and Multi-Disease Cardiac Segmentation (M&Ms) Challenge, which was recently organized as part of the MICCAI 2020 Conference. A total of 14 teams submitted different solutions to the problem, combining various baseline models, data augmentation strategies, and domain adaptation techniques. The obtained results indicate the importance of intensity-driven data augmentation, as well as the need for further research to improve generalizability towards unseen scanner vendors or new imaging protocols. Furthermore, we present a new resource of 375 heterogeneous CMR datasets acquired by using four different scanner vendors in six hospitals and three different countries (Spain, Canada and Germany), which we provide as open-access for the community to enable future research in the field.

A population group that is often overlooked in the recent revolution of self-tracking is the group of older people. This growing proportion of the general population is often faced with increasing health issues and discomfort. In order to come up with lifestyle advice towards the elderly, we need the ability to quantify their lifestyle, before and after an intervention. This research focuses on the task of activity recognition (AR) from accelerometer data. With that aim, we collect a substantial labelled dataset of older individuals wearing multiple devices simultaneously and performing a strict protocol of 16 activities (the GOTOV dataset, $$N=28$$ N = 28 ). Using this dataset, we trained Random Forest AR models, under varying sensor set-ups and levels of activity description granularity. The model that combines ankle and wrist accelerometers (GENEActiv) produced the best results (accuracy $$>80\%$$ > 80 % ) for 16-class classification. At the same time, when additional physiological information is used, the accuracy increased ($$>85\%$$ > 85 % ). To further investigate the role of granularity in our predictions, we developed the LARA algorithm, which uses a hierarchical ontology that captures prior biological knowledge to increase or decrease the level of activity granularity (merge classes). As a result, a 12-class model in which the different paces of walking were merged showed a performance above $$93\%$$ 93 % . Testing this 12-class model in labelled free-living pilot data, the mean balanced accuracy appeared to be reasonably high, while using the LARA algorithm, we show that a 7-class model (lying down, sitting, standing, household, walking, cycling, jumping) was optimal for accuracy and granularity. Finally, we demonstrate the use of the latter model in unlabelled free-living data from a larger lifestyle intervention study. In this paper, we make the validation data as well as the derived prediction models available to the community.

Concealed objects detection in terahertz imaging is an urgent need for public security and counter-terrorism. So far, there is no public terahertz imaging dataset for the evaluation of objects detection algorithms. This paper provides a public dataset for evaluating multi-object detection algorithms in active terahertz imaging. Due to high sample similarity and poor imaging quality, object detection on this dataset is much more difficult than on those commonly used public object detection datasets in the computer vision field. Since the traditional hard example mining approach is designed based on the two-stage detector and cannot be directly applied to the one-stage detector, this paper designs an image-based Hard Example Mining (HEM) scheme based on RetinaNet. Several state-of-the-art detectors, including YOLOv3, YOLOv4, FRCN-OHEM, and RetinaNet, are evaluated on this dataset. Experimental results show that the RetinaNet achieves the best mAP and HEM further enhances the performance of the model. The parameters affecting the detection metrics of individual images are summarized and analyzed in the experiments.

Research on pre-impact fall detection with wearable inertial sensors (detecting fall accidents prior to body-ground impacts) has grown rapidly in the past decade due to its great potential for developing an on-demand fall-related injury prevention system. However, most researchers use their own datasets to develop fall detection algorithms and rarely make these datasets publicly available, which poses a challenge to fairly evaluate the performance of different algorithms on a common basis. Even though some open datasets have been established recently, most of them are impractical for pre-impact fall detection due to the lack of temporal labels for fall time and limited types of motions. In order to overcome these limitations, in this study, we proposed and publicly provided a large-scale motion dataset called “KFall,” which was developed from 32 Korean participants while wearing an inertial sensor on the low back and performing 21 types of activities of daily living and 15 types of simulated falls. In addition, ready-to-use temporal labels of the fall time based on synchronized motion videos were published along with the dataset. Those enhancements make KFall the first public dataset suitable for pre-impact fall detection, not just for post-fall detection. Importantly, we have also developed three different types of latest algorithms (threshold based, support-vector machine, and deep learning), using the KFall dataset for pre-impact fall detection so that researchers and practitioners can flexibly choose the corresponding algorithm. Deep learning algorithm achieved both high overall accuracy and balanced sensitivity (99.32%) and specificity (99.01%) for pre-impact fall detection. Support vector machine also demonstrated a good performance with a sensitivity of 99.77% and specificity of 94.87%. However, the threshold-based algorithm showed relatively poor results, especially the specificity (83.43%) was much lower than the sensitivity (95.50%). The performance of these algorithms could be regarded as a benchmark for further development of better algorithms with this new dataset. This large-scale motion dataset and benchmark algorithms could provide researchers and practitioners with valuable data and references to develop new technologies and strategies for pre-impact fall detection and proactive injury prevention for the elderly.

We construct a public dataset for WiFi-based Activity Recognition named WiAR with sixteen activities operated by ten volunteers in three indoor environments. It aims to provide public signal data for researchers to reduce the cost of collected signal data and conveniently evaluate the performance of WiFi-based human activity recognition in different domains. First, we introduce the basic knowledge of WiFi signals regarding RSSI, CSI, and wireless hardware. Second, we explain the characteristics of WiAR dataset in terms of activities types, data format, data acquisition ways, and influence factors. Third, the proposed framework can estimate the quality of the shared signal data provided by other peers. Finally, we select and use five classification algorithms and two deep learning algorithms to evaluate the performance of WiAR dataset on human activity recognition. The results show that the accuracy of WiAR dataset is higher than 80% using machine learning algorithms and 90% using deep learning algorithms in different indoor environments.

Colorectal cancer is one of the world leading death causes. Fortunately, an early diagnosis allows for effective treatment, increasing the survival rate. Deep learning techniques have shown their utility for increasing the adenoma detection rate at colonoscopy, but a dataset is usually required so the model can automatically learn features that characterize the polyps. In this work, we present the PICCOLO dataset, that comprises 3433 manually annotated images (2131 white-light images 1302 narrow-band images), originated from 76 lesions from 40 patients, which are distributed into training (2203), validation (897) and test (333) sets assuring patient independence between sets. Furthermore, clinical metadata are also provided for each lesion. Four different models, obtained by combining two backbones and two encoder&ndash, decoder architectures, are trained with the PICCOLO dataset and other two publicly available datasets for comparison. Results are provided for the test set of each dataset. Models trained with the PICCOLO dataset have a better generalization capacity, as they perform more uniformly along test sets of all datasets, rather than obtaining the best results for its own test set. This dataset is available at the website of the Basque Biobank, so it is expected that it will contribute to the further development of deep learning methods for polyp detection, localisation and classification, which would eventually result in a better and earlier diagnosis of colorectal cancer, hence improving patient outcomes.

This paper reports on a benchmark dataset acquired with a brain–computer interface (BCI) system based on the rapid serial visual presentation (RSVP) paradigm. The dataset consists of 64-channel electroencephalogram (EEG) data from 64 healthy subjects (sub1,…, sub64) while they performed a target image detection task. For each subject, the data contained two groups (“A” and “B”). Each group contained two blocks, and each block included 40 trials that corresponded to 40 stimulus sequences. Each sequence contained 100 images presented at 10 Hz (10 images per second). The stimulus images were street-view images of two categories: target images with human and non-target images without human. Target images were presented randomly in the stimulus sequence with a probability of 1∼4%. During the stimulus presentation, subjects were asked to search for the target images and ignore the non-target images in a subjective manner. To keep all original information, the dataset was the raw continuous data without any processing. On one hand, the dataset can be used as a benchmark dataset to compare the algorithms for target identification in RSVP-based BCIs. On the other hand, the dataset can be used to design new system diagrams and evaluate their BCI performance without collecting any new data through offline simulation. Furthermore, the dataset also provides high-quality data for characterizing and modeling event-related potentials (ERPs) and steady-state visual evoked potentials (SSVEPs) in RSVP-based BCIs. The dataset is freely available from http://bci.med.tsinghua.edu.cn/download.html.

International audience; Driver distraction and fatigue have become one of the leading causes of severe traffic accidents. Hence, driver inattention monitoring systems are crucial. Even with the growing development of advanced driver assistance systems and the introduction of third-level autonomous vehicles, this task is still trending and complex due to challenges such as the illumination change and the dynamic background. To reliably compare and validate driver inattention monitoring methods, a limited number of public datasets are available. In this paper, we put forward a public, well-structured and complete dataset, named Multiview, Multimodal and Multispectral Driver Action Dataset (3MDAD). The dataset is mainly composed of two sets: the first one recorded in daytime and the second one at nighttime. Each set consists of two synchronized data modalities, both from frontal and side views. More than 60 drivers are asked to execute 16 in-vehicle actions under a wide range of naturalistic driving settings. In contrast to other public datasets, 3MDAD presents multiple modalities, spectrums and views under different time and weather conditions. To highlight the utility of our dataset, we independently analyze the driver action recognition results adapted to each modality and those obtained of several combinations of modalities.

International audience; A reliable environment perception is a crucial task for autonomous driving, especially in dense traffic areas. Recent improvements and breakthroughs in scene understanding for intelligent transportation systems are mainly based on deep learning and the fusion of different modalities. In this context, we introduce OLIMP: A heterOgeneous Multimodal Dataset for Advanced EnvIronMent Perception. This is the first public, multimodal and synchronized dataset that includes UWB radar data, acoustic data, narrow-band radar data and images. OLIMP comprises 407 scenes and 47,354 synchronized frames, presenting four categories: pedestrian, cyclist, car and tram. The dataset includes various challenges related to dense urban traffic such as cluttered environment and different weather conditions. To demonstrate the usefulness of the introduced dataset, we propose a fusion framework that combines the four modalities for multi object detection. The obtained results are promising and spur for future research.

Existing accelerometer-based human activity recognition (HAR) benchmark datasets that were recorded during free living suffer from non-fixed sensor placement, the usage of only one sensor, and unreliable annotations. We make two contributions in this work. First, we present the publicly available Human Activity Recognition Trondheim dataset (HARTH). Twenty-two participants were recorded for 90 to 120 min during their regular working hours using two three-axial accelerometers, attached to the thigh and lower back, and a chest-mounted camera. Experts annotated the data independently using the camera’s video signal and achieved high inter-rater agreement (Fleiss’ Kappa =0.96). They labeled twelve activities. The second contribution of this paper is the training of seven different baseline machine learning models for HAR on our dataset. We used a support vector machine, k-nearest neighbor, random forest, extreme gradient boost, convolutional neural network, bidirectional long short-term memory, and convolutional neural network with multi-resolution blocks. The support vector machine achieved the best results with an F1-score of 0.81 (standard deviation: ±0.18), recall of 0.85±0.13, and precision of 0.79±0.22 in a leave-one-subject-out cross-validation. Our highly professional recordings and annotations provide a promising benchmark dataset for researchers to develop innovative machine learning approaches for precise HAR in free living.

Automatic fall detection is a very active research area, which has grown explosively since the 2010s, especially focused on elderly care. Rapid detection of falls favors early awareness from the injured person, reducing a series of negative consequences in the health of the elderly. Currently, there are several fall detection systems (FDSs), mostly based on predictive and machine-learning approaches. These algorithms are based on different data sources, such as wearable devices, ambient-based sensors, or vision/camera-based approaches. While wearable devices like inertial measurement units (IMUs) and smartphones entail a dependence on their use, most image-based devices like Kinect sensors generate video recordings, which may affect the privacy of the user. Regardless of the device used, most of these FDSs have been tested only in controlled laboratory environments, and there are still no mass commercial FDS. The latter is partly due to the impossibility of counting, for ethical reasons, with datasets generated by falls of real older adults. All public datasets generated in laboratory are performed by young people, without considering the differences in acceleration and falling features of older adults. Given the above, this article presents the eHomeSeniors dataset, a new public dataset which is innovative in at least three aspects: first, it collects data from two different privacy-friendly infrared thermal sensors, second, it is constructed by two types of volunteers: normal young people (as usual) and performing artists, with the latter group assisted by a physiotherapist to emulate the real fall conditions of older adults, and third, the types of falls selected are the result of a thorough literature review.

Modern agriculture is facing unique challenges in building a sustainable future for food production, in which the reliable detection of plantation threats is of critical importance. The breadth of existing information sources, and their equivalent sensors, can provide a wealth of data which, to be useful, must be transformed into actionable knowledge. Approaches based on Information Communication Technologies (ICT) have been shown to be able to help farmers and related stakeholders make decisions on problems by examining large volumes of data while assessing multiple criteria. In this paper, we address the automated identification (and count the instances) of the major threat of olive trees and their fruit, the Bactrocera Oleae (a.k.a. Dacus) based on images of the commonly used McPhail trap’s contents. Accordingly, we introduce the “Dacus Image Recognition Toolkit„ (DIRT), a collection of publicly available data, programming code samples and web-services focused at supporting research aiming at the management the Dacus as well as extensive experimentation on the capability of the proposed dataset in identifying Dacuses using Deep Learning methods. Experimental results indicated performance accuracy (mAP) of 91.52% in identifying Dacuses in trap images featuring various pests. Moreover, the results also indicated a trade-off between image attributes affecting detail, file size and complexity of approaches and mAP performance that can be selectively used to better tackle the needs of each usage scenario.

During the previous years, the demand for producing the quality of software has been quickly increased. In this paper, Bayesian Regularization (BR) technique has been used for finding the software faults before the testing process. This technique helps us to reduce the cost of software testing which reduces the cost of the software project. The basic purpose of BR technique is to minimizes a combination of squared errors and weights, and then determine the correct combination so as to produce an efficient network.BR Technique algorithm based neural network tool is used for finding the results on the given public dataset. The accuracy of BR algorithm based neural network has been compared with Levenberg-Marquardt(LM) algorithm and Back Propagation (BPA) algorithm for finding the software defects. Our results signify that the software fault prediction model using BR technique provide better accuracy than Levenberg-Marquardt (LM) algorithm and Back Propagation (BPA) algorithm.