Single-shot Detection Research Articles

Detection of The Deaf Signal Language Using The Single Shot Detection (SSD) Method

Sign Language is a language that prioritizes manual communication, body language, and lip movements, instead of sound, to communicate. Deaf people are the main group who use this language, usually by combining hand shape, orientation and movement of the hands, arms, and body, and facial expressions to express their thoughts. Therefore, the researcher created an image recognition program in sign language using the Single Shot Detection (SSD) method, which is a convolution activity by combining several layers of preparation, by utilizing several components that move together and are motivated by a biological sensory system. The letters used in making sign language programs use the letters of the alphabet (az). This sign language detection programming that runs on the Google Collaboratory application

Coherent Raman spectroscopy on hydrogen with in-situ generation, in-situ use, and in-situ referencing of the ultrabroadband excitation.

Time-resolved spectroscopy can provide valuable insights in hydrogen chemistry, with applications ranging from fundamental physics to the use of hydrogen as a commercial fuel. This work represents the first-ever demonstration of in-situ femtosecond laser-induced filamentation to generate a compressed supercontinuum behind a thick optical window, and its in-situ use to perform femtosecond/picosecond coherent Raman spectroscopy (CRS) on molecular hydrogen (H2). The ultrabroadband coherent excitation of Raman active molecules in measurement scenarios within an enclosed space has been hindered thus far by the window material imparting temporal stretch to the pulse. We overcome this challenge and present the simultaneous single-shot detection of the rotational H2 and the non-resonant CRS spectra in a laminar H2/air diffusion flame. Implementing an in-situ referencing protocol, the non-resonant spectrum measures the spectral phase of the supercontinuum pulse and maps the efficiency of the ultrabroadband coherent excitation achieved behind the window. This approach provides a straightforward path for the implementation of ultrabroadband H2 CRS in enclosed environment such as next-generation hydrogen combustors and reforming reactors.

Accurate Bounding Box Prediction for Single-Shot Object Detection

Accurate single-shot object detection is an extremely challenging task in real environments because of complex scenes, occlusion, ambiguities, blur, and shadow, i.e., these factors are called uncertainty problem. It leads to unreliable labeling of bounding box annotation and makes detectors arduous to learn bounding box localization. Previous methods viewed the ground truth box coordinates as a rigid distribution omitting localization uncertainty in real datasets. This article proposes a novel bounding box encoding algorithm integrated into the single-shot detector (BBENet) to consider the flexible distribution of bounding box localization. First, discretized ground truth labels are generated by decomposing each object’s boundary into multiple boundaries. The new representation of ground truth boxes is more arbitrary and flexible to cover any case of complex scenes. During training, the detector directly learns discretized box locations instead of continuous domain. Second, the bounding box encoding algorithm reorganizes bounding box predictions to be more accurate. Furthermore, another problem in existing methods is inconsistency in estimating detection quality. The single-shot detection consists of classification and localization tasks, but the popular detectors consider the classification score as the final detection quality. Thus, it lacks localization quality and hinders the overall performance because both tasks have a positive correlation. To overcome this problem, BBENet introduces detection quality by combining the localization and classification quality to rank detection during nonmaximum suppression. The localization quality is computed based on how uncertain the predicted boxes are, which is a new perspective in detection literature. The proposed BBENet is evaluated on three benchmark datasets, i.e., MS-COCO, Pascal VOC, and CrowdHuman. Without bells and whistles, BBENet outperforms the existing methods by a large margin with comparable speed, achieving the state-of-the-art single-shot detector.

Internal defects detection and classification in hollow cylindrical surfaces using single shot detection and MobileNet

Surface defect inspection, detection, and classification in hollow cylindrical surfaces such as pipes and barrels have a significant impact on the structural integrity of various industrial products. Regular inspection and identification of the faults reduces the likelihood of faults’ aggravation, limits the damaging effects, and increases the product life. However, most of the defect detection algorithms for cylindrical surfaces rely heavily on handcrafted feature extraction limiting the ability to recognize the defects effectively. This research work proposes an image processing-based automatic defect detection and classification approach for cylindrical hollow surfaces. The proposed system uses a single shot multi-box detection (SSD) algorithm for localization and a customized lightweight deep convolutional neural network as a backbone network to classify defects generally found in industrial pipes and gun barrels. First, the image dataset is acquired from a real-time working environment using an indigenously developed borescope featuring a rotating camera and special hardware features. Later, the bounding boxes are calculated using extracted features to localize defects with SSD which takes a single shot to detect multiple objects within the image. Finally, the defected regions are classified into five classes of commonly found issues of pitting, chipping, rusting, dirt, and thermal cracking by utilizing deep learning architecture of 53 layers. It is found that the proposed approach can indicate the exact location of the classified defect in terms of angle and distance from a reference point. Also, the proposed method improves the detection and classification accuracy significantly compared to other existing methods. To encourage the development and evaluation of new methods for cylindrical surface defect detection, the dataset is also made publicly available.

V-cavity stabilized quantum cascade laser-based cavity ringdown spectroscopy for rapid detection of radiocarbon below natural abundance

Mid-infrared laser absorption spectroscopy utilizing a high-finesse optical cavity enables high precision trace analysis of gas molecules. In particular, optical detection of radiocarbon (14C) based on cavity ringdown spectroscopy using a quantum cascade laser (QCL) is gaining attention as an alternative to accelerator mass spectrometry. This paper reports a compact-packaged narrow-linewidth QCL system utilizing resonant optical feedback from an external V-shaped cavity. Based on frequency noise analysis, the derived laser linewidth is 44 kHz for 100 μs integration time with the capability to perform seamless frequency scanning around 10 GHz. We installed this laser system within a table-top cavity ringdown spectrometer for 14CO2. A single-shot detection limit of 1.2 × 10−9 cm−1 Hz−1/2 leading to a detectable abundance evaluated from a noise analysis of 0.2 in fraction modern 14C for a 10-s averaging time was achieved. This capability of rapid analysis for 14CO2 is suitable for various applications requiring trace 14C analysis.

Computer vision in aquaculture: a case study of juvenile fish counting

ABSTRACT In aquaculture breeding or production programmes, counting juvenile fish represents a considerable cost in terms of the human hours needed. In this study, we explored the use of two state-of-the-art machine learning architectures (Single Shot Detection, hereafter SSD and Faster Regions with convolutional neural networks, hereafter Faster R-CNN) to augment a manual image-based juvenile fish counting method for the Australasian snapper (Chrysophrys auratus) bred at The New Zealand Institute for Plant and Food Research Limited. We tested model accuracy after tuning for confidence thresholds and non-maximal suppression overlap parameters, and implementing a bias correction using a Poisson regression model. Validation of image data showed that after tuning, bias-corrected SSD and Faster R-CNN models had mean absolute percent errors (MAPE) of less than 10%, with SSD having MAPE of less than 5%. Comparison of the results with those from manual counts showed that, while manual counts are slightly more accurate (MAPE = 1.56), the machine learning methods allow for more rapid assessment of counts and thus facilitating a higher throughput. This work represents a first step for deploying machine learning applications to an existing real-life aquaculture scenario and provides a useful starting point for further developments, such as real-time counting of fish or collecting additional phenotypic data from the source images.

Comparison of Single-Shot and Two-Shot Deep Neural Network Models for Whitefly Detection in IoT Web Application

In this study, we have compared YOLOv4, a single-shot detector to Faster-RCNN, a two-shot detector to detect and classify whiteflies on yellow-sticky tape (YST). An IoT remote whitefly monitoring station was developed and placed in a whitefly rearing room. Images of whiteflies attracted to the trap were recorded 2× per day. A total of 120 whitefly images were labeled using labeling software and split into a training and testing dataset, and 18 additional yellow-stick tape images were labeled with false positives to increase the model accuracy from remote whitefly monitors in the field that created false positives due to water beads and reflective light on the tape after rain. The two-shot detection model has two stages: region proposal and then classification of those regions and refinement of the location prediction. Single-shot detection skips the region proposal stage and yields final localization and content prediction at once. Because of this difference, YOLOv4 is faster but less accurate than Faster-RCNN. From the results of our study, it is clear that Faster-RCNN (precision—95.08%, F-1 Score—0.96, recall—98.69%) achieved a higher level of performance than YOLOv4 (precision—71.77%, F-1 score—0.83, recall—73.31%), and will be adopted for further development of the monitoring station.

Lightweight jet reconstruction and identification as an object detection task

We apply object detection techniques based on deep convolutional blocks to end-to-end jet identification and reconstruction tasks encountered at the CERN large hadron collider (LHC). Collision events produced at the LHC and represented as an image composed of calorimeter and tracker cells are given as an input to a Single Shot Detection network. The algorithm, named PFJet-SSD performs simultaneous localization, classification and regression tasks to cluster jets and reconstruct their features. This all-in-one single feed-forward pass gives advantages in terms of execution time and an improved accuracy w.r.t. traditional rule-based methods. A further gain is obtained from network slimming, homogeneous quantization, and optimized runtime for meeting memory and latency constraints of a typical real-time processing environment. We experiment with 8-bit and ternary quantization, benchmarking their accuracy and inference latency against a single-precision floating-point. We show that the ternary network closely matches the performance of its full-precision equivalent and outperforms the state-of-the-art rule-based algorithm. Finally, we report the inference latency on different hardware platforms and discuss future applications.

Terahertz-wave decoding of femtosecond extreme-ultraviolet light pulses

In recent years, femtosecond extreme-ultraviolet (XUV) and x-ray pulses from free-electron lasers have developed into important probes to monitor processes and dynamics in matter on femtosecond-time and angstrom-length scales. With the rapid progress of versatile ultrafast x-ray spectroscopy techniques and more sophisticated data analysis tools, accurate single-pulse information on the arrival time, duration, and shape of the probing x-ray and XUV pulses becomes essential. Here, we demonstrate that XUV pulses can be converted into terahertz electromagnetic pulses using a spintronic terahertz emitter. We observe that the duration, arrival time, and energy of each individual XUV pulse is encoded in the waveform of the associated terahertz pulses, and thus can be readily deduced from single-shot terahertz time-domain detection.

Single-shot terahertz polarization detection based on terahertz time-domain spectroscopy

This paper presents a novel design for single-shot terahertz polarization detection based on terahertz time-domain spectroscopy (THz-TDS). Its validity has been confirmed by comparing its detection results with those of the THz common-path spectral interferometer through two separate measurements for the orthogonal components. Our results also show that its detection signal-to-noise ratios (SNRs) are obviously superior to those of the 45° optical bias THz-TDS by electro-optical sampling due to its operation on common-path spectral interference rather than the polarization-sensitive intensity modulation. The setup works without need of any optical scan, which does not only save time, but also efficiently avoids the disturbances from the fluctuations of the system and environment. Its single-shot mode allows it to work well for the applications with poor or no repeatability.

Face Detection by Measuring Thermal Value to Avoid Covid-19

In this study, custom Single Shot Detection (SSD) was used to detect infected people faces because of Covid-19. It has been suggested to determine around the eyes area where the body temperature of the person most accurate by determining the facial landmarks with Ensemble of Regression Trees (ERT) model on these detected faces. Finally, the thermal value was measured from around the eyes area in a non-contact way using sensor fusion. As a result of the analyzes made, it was observed that the proposed system gave results close to the different measurement methods

An Improved Kernel Correlation Filter Tracking Combined with Mobilenet SSD

<p>This article mainly solves the problems that exist when using the Kernel Correlation Filter (KCF) for tracking in complex scenarios. To make the algorithm suitable for target tracking under complex conditions such as scale changes, similar interference, and occlusion, a MobileNet SSD (Single Shot Detection) target detection combined with an improved KCF target tracking algorithm is proposed. Firstly, the MobileNet SSD is used to locate the target in the initial frame, and the location is sent to KCF for training. Secondly, aiming at the problem of scale changes, a Binary-Tree scale search strategy is proposed. In this strategy, the scale value is searched in a tree shape according to the response size, which reduces the number of scale searches. Finally, the average peak correlation energy is used for occlusion determination, and the model update strategy is improved, thereby enhancing the algorithm’s ability to track occluded targets. The results of experimental evaluation and comparison on the OTB100 and UAV123 data sets show that when the target has complex conditions such as scale changes, similar interference, occlusion, etc., the proposed algorithm performs well in mainstream related filtering algorithms. Through the quantitative and qualitative analysis of the experimental results, the effectiveness of the proposed algorithm is verified.</p> <p> </p>

Cavity-dumping a single infrared pulse from a free-electron laser for two-color pump-probe experiments.

Electromagnetic radiation in the mid- to far-infrared spectral range represents an indispensable tool for the study of numerous types of collective excitations in solids and molecules. Short and intense pulses in this terahertz spectral range are, however, difficult to obtain. While wide wavelength-tunability is easily provided by free-electron lasers, the energies of individual pulses are relatively moderate, on the order of microjoules. Here, we demonstrate a setup that uses cavity-dumping of a free-electron laser to provide single, picosecond-long pulses in the mid- to far-infrared frequency range. The duration of the Fourier-limited pulses can be varied by cavity detuning, and their energy was shown to exceed 100 µJ. Using the aforementioned infrared pulse as a pump, we have realized a two-color pump-probe setup facilitating single-shot time-resolved imaging of magnetization dynamics. We demonstrate the capabilities of the setup first on thermally induced demagnetization and magnetic switching of a GdFeCo thin film and second by showing a single-shot time-resolved detection of resonant phononic switching of the magnetization in a magnetic garnet.

Alignment control using visual servoing and mobilenet single-shot multi-box detection (SSD): a review

The concept is highly critical for robotic technologies that rely on visual feedback. In this context, robot systems tend to be unresponsive due to reliance on pre-programmed trajectory and path, meaning the occurrence of a change in the environment or the absence of an object. This review paper aims to provide comprehensive studies on the recent application of visual servoing and DNN. PBVS and Mobilenet-SSD were chosen algorithms for alignment control of the film handler mechanism of the portable x-ray system. It also discussed the theoretical framework features extraction and description, visual servoing, and Mobilenet-SSD. Likewise, the latest applications of visual servoing and DNN was summarized, including the comparison of Mobilenet-SSD with other sophisticated models. As a result of a previous study presented, visual servoing and MobileNet-SSD provide reliable tools and models for manipulating robotics systems, including where occlusion is present. Furthermore, effective alignment control relies significantly on visual servoing and deep neural reliability, shaped by different parameters such as the type of visual servoing, feature extraction and description, and DNNs used to construct a robust state estimator. Therefore, visual servoing and MobileNet-SSD are parameterized concepts that require enhanced optimization to achieve a specific purpose with distinct tools.

Vision Navigator: A Smart and Intelligent Obstacle Recognition Model for Visually Impaired Users

Vision impairment is a major challenge faced by humanity on a large scale throughout the world. Affected people find independently navigating and detecting obstacles extremely tedious. Thus, a potential solution for accurately detecting obstacles requires an integrated deployment of the Internet of Things and predictive analytics. This research introduces “Vision Navigator,” a novel framework for assisting visually impaired users in obstacle analysis and tracking so that they can move independently. An intelligent stick named “Smart-fold Cane” and sensor-equipped shoes called “Smart-alert Walker” are the main constituents of our proposed model. For object detection and classification, the stick uses a single-shot detection (SSD) mechanism, which is followed by frame generation using the recurrent neural network (RNN) model. Smart-alert Walker is a lightweight shoe that acts as an emergency unit that notifies the user regarding the presence of any obstacle within a short distance range. This intelligent obstacle detection model using the SSD-RNN approach was deployed in real time and its performance was validated in indoor and outdoor environments. The SSD-RNN model computed an optimum accuracy of 95.06% and 87.68% indoors and outdoors, respectively. The model was also evaluated in the context of users’ distance from obstacles. The proposed SSD-RNN model had an accuracy rate of 96.4% and 86.8% for close and distant obstacles, respectively, outperforming other models. Execution time for the SSD-RNN model was 4.82 s with the highest mean accuracy rate of 95.54% considering all common obstacles.

High Energy and Thermal Neutron Sensitivity of Google Tensor Processing Units

In this article, we investigate the reliability of Google’s coral tensor processing units (TPUs) to both high-energy atmospheric neutrons (at ChipIR) and thermal neutrons from a pulsed source [at equipment materials and mechanics analyzer (EMMA)] and from a reactor [at Thermal and Epithermal Neutron Irradiation Station (TENIS)]. We report data obtained with an overall fluence of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$3.41\times {10^{12}} {\mathrm {n/cm}}^{2}$ </tex-math></inline-formula> for atmospheric neutrons (equivalent to more than 30 million years of natural irradiation) and of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$7.55\times {10^{12}} {\mathrm {n/cm}}^{2}$ </tex-math></inline-formula> for thermal neutrons. We evaluate the behavior of TPUs executing elementary operations with increasing input sizes (standard convolutions or depthwise convolutions) as well as eight convolutional neural networks (CNNs) configurations (single-shot multibox detection (SSD) MobileNet v2 and SSD MobileDet, trained with COCO dataset, and Inception v4 and ResNet-50, with ILSVRC2012 dataset). We found that, despite the high error rate, most neutron-induced errors only slightly modify the convolution output and do not change the detection or classification of CNNs. By reporting details about the error model, we provide valuable information on how to design the CNNs to avoid neutron-induced events to lead to misdetections or classifications.

Geometric-phase-based shearing interferometry for broadband vortex state decoding

Given that spin and orbital angular momenta of photons have been widely investigated in optical communication and information processing systems, efficient decoding of optical vortex states using a single element is highly anticipated. In this work, a wavelength-independent holographic scheme has been proposed for total angular momentum sorting of both scalar and vector vortex states with a stationary broadband geometric-phase waveplate by means of reference-free shearing interferometry. The entangled spin and orbital angular momentum modes can be distinguished simultaneously based on the spin–orbit optical Hall effect in order to realize single-shot vortex detection. The viability of our scheme has also been demonstrated experimentally.

A method of single‐shot target detection with multi‐scale feature fusion and feature enhancement

The Single Shot MultiBox Detector (SSD) is one of the fastest detection algorithms. Although it has achieved good results in detection, it also has the problem of poor detection effect for small targets and occlusion between objects. Here, the authors propose a new target detection method called single-shot target detection with multi-scale feature fusion and feature enhancement. Here, the authors introduce multi-scale feature fusion module, feature enhancement module and efficient channel attention module, and integrate them into the detection module of the original SSD target detection algorithm to improve the ability of network feature extraction. Experimental results on pascal VOC 2007 datasets show that the proposed algorithm works well when the input size is 300 × 300, the detection speed reaches 41.7 frames per second (FPS) and the detection accuracy reaches 79.6%, which is 2.4% higher than the original SSD target detection algorithm. When the input size is 512 × 512, the detection accuracy is 81.9%, and the detection speed reaches 36.5 FPS, which is 3.2% higher than the original SSD target detection algorithm. According to the experimental results, our algorithm has a better performance when there are many objects in the image and there is occlusion.

Highly interacting machining feature recognition via small sample learning

In the area of intelligent manufacturing, recognising the interacting features on a CAD model is a critical yet challenging task as topology structures of features are damaged due to the feature interaction. Some of the learning-based feature recognition methods produce less favourable results when recognising highly interacting features, while some require a significant amount of 3D models for training, which present an increasing challenge in a real world scenario, especially whenever collecting large training data becomes too difficult and time-consuming. To this end, effective highly interacting feature recognition via small sample learning becomes bottleneck for learning-based methods. To tackle the above issue, the paper proposes a novel method named RDetNet based on single-shot refinement object detection network (RefineDet) which is capable of recognising highly interacting features with small training samples. In addition, the paper further utilises several data augmentation (DA) strategies to increase the amount of relevant 3D training models. Experiments carried out in this paper show that the proposed method yields favourable results in recognising highly interacting features by using small training samples (e.g. 32 models per class).

Development of the Osteosarcoma Lung Nodules Detection Model Based on SSD-VGG16 and Competency Comparing With Traditional Method

Osteosarcoma nodule that metastasized to the patient&#x2019;s lungs was difficult to detect due to limited cases caused by its rarity. The traditional method for finding lung nodules is manually done by radiologists by looking at CT-scanned images. As a result, the error rate for reading lung metastasized nodules ranged from 29 to 42 percent, while the permissible mistake rate for reading should be less than 29 percent. Advanced computer-aid techniques such as image processing and machine learning can help doctors to identify the Osteosarcoma lung nodules easier and more accurately. Convolutional Neural Networks (CNNs) are promising techniques since they could be trained by experienced radiologists. Nodule location and size information was critical for treatments that were obtained by object detector CNNs models. In this research, the Single Shot Detection (SSD) framework combined with the VGG16 backbone, SSD-VGG16, was implemented to obtain bounding box locations and sizes when each box represents one Osteosarcoma nodule with the confidence score. The SSD-VGG16 was selected due to its superior performance. The patient&#x2019;s CT-scanned images dataset collected from 202 patient cases was provided by Lerdsin hospital and used for training and validating the SSD-VGG16 model. The trained SSD-VGG16 model was trained based on two loss functions which are class confidence and location losses. Then, the trained model experimented with unseen CT-scanned images. The performance scores were calculated. The Result was analyzed and concluded. Finally, SSD-VGG16 shows the ability to detect and locate the nodules efficiently and has less error compared to the traditional method.