Fingertip detection and tracking for recognition of air-writing in videos
تشخیص و ردیابی اثر انگشت برای تشخیص هوا-نوشتاری در فیلم ها-2019
Air-writing is the process of writing characters or words in free space using finger or hand movements without the aid of any hand-held device. In this work, we address the problem of mid-air finger writ- ing using web-cam video as input. In spite of recent advances in object detection and tracking, accurate and robust detection and tracking of the fingertip remains a challenging task, primarily due to small di- mension of the fingertip. Moreover, the initialization and termination of mid-air finger writing is also challenging due to the absence of any standard delimiting criterion. To solve these problems, we pro- pose a new writing hand pose detection algorithm for initialization of air-writing using the Faster R-CNN framework for accurate hand detection followed by hand segmentation and finally counting the num- ber of raised fingers based on geometrical properties of the hand. Further, we propose a robust finger- tip detection and tracking approach using a new signature function called distance-weighted curvature entropy. Finally, a fingertip velocity-based termination criterion is used as a delimiter to mark the com- pletion of the air-writing gesture. Experiments show the superiority of the proposed fingertip detection and tracking algorithm over state-of-the-art approaches giving a mean precision of 73.1% while achiev- ing real-time performance at 18.5 fps, a condition which is of vital importance to air-writing. Character recognition experiments give a mean accuracy of 96.11% using the proposed air-writing system, a result which is comparable to that of existing handwritten character recognition systems.
Keywords: Air-writing | Hand pose detection | Fingertip detection and tracking | Handwritten character recognition | Human-computer interaction (HCI)
Automatic detection, localization and segmentation of nano-particles with deep learning in microscopy images
تشخیص خودکار ، بومی سازی و تقسیم نانو ذرات با یادگیری عمیق در تصاویر میکروسکوپی-2019
With the growing amount of high resolution microscopy images automatic nano-particle detection, shape analysis and size determination have gained importance for providing quantitative support that gives important information for the evaluation of the material. In this paper, we present a new method for detection of nanoparticles and determination of their shapes and sizes simultaneously with deep learning. The proposed method employs multiple output convolutional neural networks (MO-CNN) and has two outputs: first is the detection output that gives the locations of the particles and the other one is the segmentation output for providing the boundaries of the nano-particles. The final sizes of particles are determined with the modified Hough algorithm that runs on the segmentation output. The proposed method is tested and evaluated on a dataset containing 17 TEM images of Fe3O4 and silica coated nano-particles. Also, we compared these results with U-net algorithm which is a popular deep learning method. The experiments showed that the proposed method has 98.23% accuracy for detection and 96.59% accuracy for segmentation of nano-particles.
Keywords: Nano-particle | Deep learning | Object detection | MO-CNN | Hough transform
Security-aware multi-objective optimization of distributed reconfigurable embedded systems
بهینه سازی چند هدفه امنیت آگاه سیستم های جاسازی شده قابل تنظیم مجدد توزیع شده-2019
Distributed embedded systems are increasingly prevalent in numerous applications, and with pervasive network access within these systems, security is also a critical design concern. We present a modeling and optimization framework for distributed embedded systems incorporating heterogeneous resources, including single core processor, asymmetric multicore processors, and FPGAs. A dataflow-based modeling framework for streaming applications integrates models for computational latency, cryptographic security levels, communication latency, and power consumption. We utilize a multi-objective genetic optimization algorithm to optimize security subject to constraints for energy consumption and minimum security level. The presented methodology is evaluated using a video-based object detection and tracking application considering several distributed heterogeneous embedded systems architectures.
Keywords: Distributed embedded systems | Security | Co-design modeling | Dynamic optimization | Design space exploration | Penalty functions
Automatic staging model of heart failure based on deep learning
مدل مرحله بندی خودکار نارسایی قلبی مبتنی بر یادگیری عمیق-2019
Heart failure (HF) is a disease that is harmful to human health. Recent advances in machine learningyielded new techniques to train deep neural networks, which resulted in highly successful applica-tions in many pattern recognition tasks such as object detection and speech recognition. To improve thediagnostic accuracy of HF staging, this study evaluates the performance of deep learning-based modelson combined features for its categorization. We proposed a novel deep convolutional neural network-Recurrent neural network (CNN-RNN) model for automatic staging of heart failure diseases in real-timeand dynamically. We employed the data segmentation and data augmentation pre-processing datasetto make the classification performance of the proposed architecture better. Specifically, this paper useconvolutional neural network (CNN) as a feature extractor instead of training the entire network toextract the characteristics of the electrocardiogram (ECG) signals and form a feature set. We combine theabove feature set with other clinical features, feed the combined features to RNN for classification, andfinally obtain 5 classification results. Experiments shows that the CNN-RNN model proposed in this paperachieved an accuracy of 97.6%, the sensitivity of 96.3%, specificity of 97.4% and proportion of 97.1% fortwo seconds of ECG segments. We obtained an accuracy, sensitivity, specificity and proportion of 96.2%,96.9%, 95.7%, and 94.3% respectively for five seconds of ECG duration. The model can be used as an aid tohelp clinicians confirm their diagnosis.
Keywords:Heart failure | Staging model | Deep learning | Deep CNN-RNN model
تشخیص و شناسایی ترافیک بر اساس شبکههای پیچشی هرمی
سال انتشار: 2019 - تعداد صفحات فایل pdf انگلیسی: 11 - تعداد صفحات فایل doc فارسی: 28
با توسعه فناوری بدونراننده، ما به شدت نیاز به روشی برای درک صحنههای ترافیکی داریم. با این حال هنوز شناسایی علائم راهنمایی و رانندگی به دلیل مقیاس کوچک این نشانهها در تصاویر جهان واقعی، وظیفهای دشوار است. در سناریوهای پیچیده برخی علائم راهنمایی و رانندگی به دلیل شرایط آب و هوایی بسیار بد و شرایط نورپردازی میتواند بسیار اغفالکننده باشد. برای پیادهسازی یک سیستم تشخیص و شناسایی جامعتر ما یک شبکه دو مرحلهای را توسعه میدهیم. در مرحله پیشنهاد ناحیه، ما یک معماری عرمی ویژگی عمیق را با اتصالات جانبی به کار میگیریم که سبب میشود ویژگیهای معنایی شی کوچک حساستر شوند. در مرحله طبقهبندی شبکه پیچیشی که به شکل متراکم متصل شده است به منظور تقویت انتقال و تسهیم ویژگی مورد استفاده واقع شده است که این شبکه منجر به طبقهبندی دقیقتر با تعداد پارامترهای کمتر خواهد شد. ما بر روی بنچمارک تشخیص GTSDB و همچنین بر روی بنچمارک چالش برانگیز k100 Tsinghua-Tencent نیز آزمایش کردیم که برای اکثر شبکههای سنتی بسیار مشکل است. آزمایشات نشان میدهند که روش پیشنهادی ما عملکردی بسیار عالی را کسب میکند و از سایر جدیدترین روشها نیز بهتر است. پیادهسازی کد منبع در آدرس روبرو در دسترس است: https://github.com/derderking/Traffic-Sign.
کلیدواژهها: نشانه ترافیک | تشخیص شی | هرم ویژگی.
|مقاله ترجمه شده|
A review: Deep learning for medical image segmentation using multi-modality fusion
یک مرور: یادگیری عمیق برای تقسیم بندی تصویر پزشکی با استفاده از همجوشی چند مدلی-2019
Multi-modality is widely used in medical imaging, because it can provide multiinformation about a target (tumor, organ or tissue). Segmentation using multimodality consists of fusing multi-information to improve the segmentation. Recently, deep learning-based approaches have presented the state-of-the-art performance in image classification, segmentation, object detection and tracking tasks. Due to their self-learning and generalization ability over large amounts of data, deep learning recently has also gained great interest in multi-modal medical image segmentation. In this paper, we give an overview of deep learning-based approaches for multi-modal medical image segmentation task. Firstly, we introduce the general principle of deep learning and multi-modal medical image segmentation. Secondly, we present different deep learning network architectures, then analyze their fusion strategies and compare their results. The earlier fusion is commonly used, since it’s simple and it focuses on the subsequent segmentation network architecture. However, the later fusion gives more attention on fusion strategy to learn the complex relationship between different modalities. In general, compared to the earlier fusion, the later fusion can give more accurate result if the fusion method is effective enough. We also discuss some common problems in medical image segmentation. Finally, we summarize and provide some perspectives on the future research.
Keywords: Deep learning | Medical image segmentation | Multi-modality fusion | Review
Diagnostic assessment of deep learning algorithms for diabetic retinopathy screening
ارزیابی تشخیصی الگوریتم های یادگیری عمیق برای غربالگری رتینوپاتی دیابتی-2019
Diabetic retinopathy (DR), the leading cause of blindness for working-age adults, is gen- erally intervened by early screening to reduce vision loss. A series of automated deep- learning-based algorithms for DR screening have been proposed and achieved high sensi- tivity and specificity ( > 90%). However, these deep learning models do not perform well in clinical applications due to the limitations of the existing publicly available fundus im- age datasets. In order to evaluate these methods in clinical situations, we collected 13,673 fundus images from 9598 patients. These images were divided into six classes by seven graders according to image quality and DR level. Moreover, 757 images with DR were se- lected to annotate four types of DR-related lesions. Finally, we evaluated state-of-the-art deep learning algorithms on collected images, including image classification, semantic seg- mentation and object detection. Although we obtain an accuracy of 0.8284 for DR classi- fication, these algorithms perform poorly on lesion segmentation and detection, indicating that lesion segmentation and detection are quite challenging. In summary, we are provid- ing a new dataset named DDR for assessing deep learning models and further exploring the clinical applications, particularly for lesion recognition.
Keywords: Diabetic retinopathy | Fundus image | Deep learning | Image classification | Semantic segmentation
Foveated ghost imaging based on deep learning
تصویربرداری از خیال مبتنی بر یادگیری عمیق-2019
Ghost imaging is an unconventional imaging mechanism that utilizes the high-order correlation to reconstruct object’s image. Limited by the maximum refresh rate of DMD or SLM, the sampling efficiency of ghost imaging has been a major obstacle for practical application. In this paper, foveated ghost imaging based on deep learning (DPFGI) is proposed to generate non-uniform resolution speckle patterns according to the object detection results as the fovea point. We combine foveated speckle pattern inspired by the human visual system with GAN-based ghost imaging object detection system to realize selecting the region of interest for foveated imaging intelligently. The simulation and experimental results show that DPFGI can detect objects in undersampled images with higher accuracy and achieve higher PSNR in the fovea region compared with uniform-resolution ghost imaging, which opens new perspectives for more intelligent ghost imaging.
Keywords: Ghost imaging | Foveated imaging | Deep learning | Object detection
Computer vision detection of foreign objects in walnuts using deep learning
تشخیص بینایی ماشین از اشیاء خارجی در گردو با استفاده از یادگیری عمیق-2019
Detection of foreign objects is crucial for quantitative image analysis in numerous food quality and safety inspection applications. Rapid detection of foreign objects in walnuts using computer vision still faces challenge due to the irregular shapes and complex features of foreign objects. Some detection methods require application specific transforms, expertly designed constraints and model parameters, and have limited detection performance due to their maintenance costs. In recent years, deep learning has become a focus in different research fields, because methods based on deep learning are able to directly learn features from training data. In this study, we apply two different convolutional neural network structures to walnut images for automatically segmenting images and detecting different-sized natural foreign objects (e.g., flesh leaf debris, dried leaf debris and gravel dust) and man-made foreign objects (e.g., paper scraps, packing material, plastic scraps and metal parts). The proposed deep-learning method is simpler because it avoids extracting features manually, and overcomes the conglomeration phenomenon between walnuts and foreign objects in actual images. The proposed method is able to correctly segment 99.5% of the object regions in the 101 test images and to correctly classify 95% of the foreign objects in the 277 validation images. The segmentation and detection processing time of each image was less than 50 ms. Future work will focus on deep learning using multi-waveband imaging hardware and fast on-line inspection control for the equipment and robots.
Keywords: Computer vision | Walnuts | Deep learning | Foreign object detection
Automatic object detection using dynamic time warping on ground penetrating radar signals
ردیابی خودکار شی با استفاده از چرخش زمانی پویا در سیگنالهای رادار نافذ در زمین-2019
Ground Penetrating Radar (GPR) is a widely used non-destructive method in buried object detection. However, online, automatic, and accurate location and depth estimation methods using GPR are still un- der development. In this article, a cutting-edge expert system is proposed that compares signals from newly scanned locations to a target-free accumulated reference signal and computes a dissimilarity mea- sure using Dynamic Time Warping (DTW). By setting a threshold on DTW values and monitoring them online, a significant deviation of the DTW values from the reference signal is detected prior to reaching an object. A potential burial site is therefore automatically detected without having a complete GPR scan which is a huge advantage compared to existing methods. Following the scanning process and investi- gating the potential burial site, location and depth of multiple buried objects is estimated automatically and highly accurate. The fully-automated analytics eliminate the need of expert operators in estimating spatial burial locations and perform accurately even on noisy media. Statistical proofs are provided that support the validity of the developed expert system in theory. Moreover, the analytics run in real-time that is plausible for on-site applications
Keywords: Ground Penetrating Radar signals | Dynamic Time Warping | Sequential confidence intervals | Control process | Object detection