Large-scale Detection Research Articles

Deep Learning Approach for Paddy Field Detection Using Labeled Aerial Images: The Case of Detecting and Staging Paddy Fields in Central and Southern Taiwan

Detecting and mapping paddy fields in Taiwan’s agriculture is crucial for managing agricultural production, predicting yields, and assessing damages. Although researchers at the Taiwan Agricultural Research Institute currently use site surveys to identify rice planting areas, this method is time-consuming. This study aimed to determine the optimal band combinations and vegetation index for accurately detecting paddy fields during various phenological stages. Additionally, the Mask RCNN instance segmentation model in the ArcGIS Pro software was employed to enhance the effectiveness of detecting and segmenting paddy fields in aerial images. This study utilized aerial images collected from 2018 to 2019 covering Changhua, Yunlin, Chiayi, and Tainan in central and southern Taiwan, with a label file comprising four categories of rice growing, ripening, harvested stage, and other crops. To create different image datasets, the image pre-processing stage involved modifying band information using different vegetation indices, including NDVI, CMFI, DVI, RVI, and GRVI. The resolution of the training image chips was cropped to 550 by 550 pixels. After the model training process, the study found that the ResNet-50 backbone performed better than the ResNet-101, and the RGB + DVI image dataset achieved the highest mean average precision of 74.01%. In addition, the model trained on the RGB + CMFI image dataset was recommended for detecting paddy fields in the rice growing stage, RGB + NIR for the rice ripening stage, and RGB + GRVI for the rice harvested stage. These models exhibit Dice coefficients of 79.59%, 89.71%, and 87.94%, respectively. The detection and segmentation results can improve the efficiency of rice production management by using different band combinations according to different rice phenological stages. Furthermore, this method can be applied to large-scale detection of other crops, improving land use survey efficiency and reducing the burden on researchers.

Curriculum learning-based strategy for low-density archaeological mound detection from historical maps in India and Pakistan

This paper presents two algorithms for the large-scale automatic detection and instance segmentation of potential archaeological mounds on historical maps. Historical maps present a unique source of information for the reconstruction of ancient landscapes. The last 100 years have seen unprecedented landscape modifications with the introduction and large-scale implementation of mechanised agriculture, channel-based irrigation schemes, and urban expansion to name but a few. Historical maps offer a window onto disappearing landscapes where many historical and archaeological elements that no longer exist today are depicted. The algorithms focus on the detection and shape extraction of mound features with high probability of being archaeological settlements, mounds being one of the most commonly documented archaeological features to be found in the Survey of India historical map series, although not necessarily recognised as such at the time of surveying. Mound features with high archaeological potential are most commonly depicted through hachures or contour-equivalent form-lines, therefore, an algorithm has been designed to detect each of those features. Our proposed approach addresses two of the most common issues in archaeological automated survey, the low-density of archaeological features to be detected, and the small amount of training data available. It has been applied to all types of maps available of the historic 1″ to 1-mile series, thus increasing the complexity of the detection. Moreover, the inclusion of synthetic data, along with a Curriculum Learning strategy, has allowed the algorithm to better understand what the mound features look like. Likewise, a series of filters based on topographic setting, form, and size have been applied to improve the accuracy of the models. The resulting algorithms have a recall value of 52.61% and a precision of 82.31% for the hachure mounds, and a recall value of 70.80% and a precision of 70.29% for the form-line mounds, which allowed the detection of nearly 6000 mound features over an area of 470,500 km2, the largest such approach to have ever been applied. If we restrict our focus to the maps most similar to those used in the algorithm training, we reach recall values greater than 60% and precision values greater than 90%. This approach has shown the potential to implement an adaptive algorithm that allows, after a small amount of retraining with data detected from a new map, a better general mound feature detection in the same map.

Multi-scale multivariate detection method for the effective impact range of hydraulic fracturing in coal seam

Hydraulic fracturing enhances coal seam permeability and reduces original gas content. Determining the effective range of hydraulic fracturing is essential in evaluating its effectiveness. This paper analyzes physical field changes during hydraulic fracturing and proposes a multi-scale multivariate detection method for determining the effective range of hydraulic fracturing in coal seams based on changes in wave velocity and resistivity due to fracture expansion.The method was validated through field experiments. The result shows that vibration wave computed tomography (VWCT) is a passive, large-scale area detection method. In contrast, the direct current (DC) method is a small-scale, local active detection method. In the hydraulic fracturing process, the microseismic (MS) events increase. After fracturing, the wave velocity of coal seam decreases significantly, indicating the formation of fractures around the fractured holes and a fracture zone. The fracturing expands the pre-existing fractures adjacent to fractured holes, forming a fracture network between them and natural fractures, leading to a decrease in wave velocity around the coal seam. The DC detection results indicate a noticeable development of pores around the fractured holes following hydraulic fracturing. The increased resistance value area is mainly concentrated on both sides of the 9–4 drilling, with an influence range of approximately 30 m, and an increased range of 100–400 Ω·m, indicating a hydraulic fracturing pore expansion range of 30 m in coal seams. Comprehensively, the maximum impact area of fracturing is the area surrounding the fractured holes, and the impact area is not uniform. The DC method test results reflect the extension of fracturing around fractured holes, while the VWCT test results represent the extension of fracturing around fractured holes and the connectivity of primary fractures, both of which include each other.

A Cross-Scale and Illumination Invariance-Based Model for Robust Object Detection in Traffic Surveillance Scenarios

Robust object detection methods in traffic surveillance scenarios often encounters challenges due to large-scale deformations and illumination variations in outdoor scenes. To enhance the tolerance of such methods against these variations, we design a cross-scale and illumination-invariant detection model (CSIM) based on the You Only Look Once (YOLO) architecture. A main cause of false detection in large-scale detection tasks is the inconsistency between various feature scales. To address this issue, we introduce an adaptive cross-scale feature fusion model to ensure the consistency of the constructed feature pyramid. To overcome the influence of uneven light, we build an illumination-invariant chromaticity space on the CSIM model, which is independent of the correlated color temperature. In addition, we adopt spatial attention modules, K-means clustering and the Mish activation function for further model optimization. The obtained experimental results show that the proposed CSIM produces excellent detection results for addressing the challenges derived from large-scale deformations and the illumination changes encountered during traffic surveillance. Compared with state-of-the-art object detection methods on public datasets, our proposed model has achieved competitive results in robust object detection tasks in traffic surveillance scenarios.

Identifying abandoned buildings in shrinking cities with mobile sensing images

The number of abandoned buildings in shrinking cities is increasing sharply, posing environment risks, threatening the safety and health of residents, affecting the real estate market, and burdening government finance. Abandoned building detection provides fundamental information for refined urban management, real estate transactions and government decision-making. However, emerging sources of data, such as satellite imagery and commercial street views, are insufficient to timely collect this fine-scale data, lacking large-scale and fine-grained detection method. Therefore, in this research, we aim to define the connotation and identification criteria of abandoned buildings, develop an effective deep learning method based on image segmentation, and detect individual abandoned buildings from large-scale mobile sensing images (MSIs) with high accuracy. The study conducted a mobile sensing campaign in a shrinking city in Northeast China, collecting 11,359 street-level images of 126.2 km of urban roads. The accuracy of the deep learning detection method was 83.8%. The study compared with the detection of commercial street view images (latest in 2015) and analyzed the dynamic changes of abandoned buildings. From 2015 to 2021, the number of abandoned buildings in the case city decreased from 102 to 50 and became more concentrated in the old city area. Our study demonstrates the feasibility of MSIs in detecting abandoned buildings and shows the enormous potential to timely detect abandoned buildings in large spatial ranges.

End-to-End Zero-Shot HOI Detection via Vision and Language Knowledge Distillation

Most existing Human-Object Interaction (HOI) Detection methods rely heavily on full annotations with predefined HOI categories, which is limited in diversity and costly to scale further. We aim at advancing zero-shot HOI detection to detect both seen and unseen HOIs simultaneously. The fundamental challenges are to discover potential human-object pairs and identify novel HOI categories. To overcome the above challenges, we propose a novel End-to-end zero-shot HOI Detection (EoID) framework via vision-language knowledge distillation. We first design an Interactive Score module combined with a Two-stage Bipartite Matching algorithm to achieve interaction distinguishment for human-object pairs in an action-agnostic manner. Then we transfer the distribution of action probability from the pretrained vision-language teacher as well as the seen ground truth to the HOI model to attain zero-shot HOI classification. Extensive experiments on HICO-Det dataset demonstrate that our model discovers potential interactive pairs and enables the recognition of unseen HOIs. Finally, our method outperforms the previous SOTA under various zero-shot settings. Moreover, our method is generalizable to large-scale object detection data to further scale up the action sets. The source code is available at: https://github.com/mrwu-mac/EoID.

Spike- and nucleocapsid-based gold colloid assay toward the development of an adhesive bandage for rapid SARS-CoV-2 immune response detection and screening

Immunoglobulin M (IgM) and immunoglobulin G (IgG) antibodies are important biomarkers used for the diagnosis and screening of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections in both symptomatic and asymptomatic individuals. These antibodies are highly specific to the spike (S) and nucleocapsid (N) proteins of the SARS-CoV-2 virus. This paper outlines the development steps of a novel hybrid (vertical-lateral-vertical) flow assay in the form of a finger-stick point-of-care device, similar to an adhesive bandage, designed for the timely detection and screening of IgM and IgG immune responses to SARS-CoV-2 infections. The assay, comprising a vertically stacked plasma/serum separation membrane, conjugate pad, and detection (readout) zone, utilizes gold nanoparticles (AuNPs) conjugated with SARS-CoV-2 S and N proteins to effectively capture IgM and IgG antibodies from a pinprick (~15 µL) of blood in just one step and provides results of no immune IgM−/IgG−, early immune IgM+/IgG−, active immune IgM+/IgG+ or immune IgM−/IgG+ in a short amount of time (minutes). The adhesive bandage-like construction is an example of the design of rapid, low-cost, disposable, and easy-to-use tests for large-scale detection and screening in households. Furthermore, the bandage can be easily adjusted and optimized to detect different viral infections as they arise by simply selecting appropriate antigens related to pandemics and outbreaks.

A large-scale data security detection method based on continuous time graph embedding framework

Graph representation learning has made significant strides in various fields, including sociology and biology, in recent years. However, the majority of research has focused on static graphs, neglecting the temporality and continuity of edges in dynamic graphs. Furthermore, dynamic data are vulnerable to various security threats, such as data privacy breaches and confidentiality attacks. To tackle this issue, the present paper proposes a data security detection method based on a continuous-time graph embedding framework (CTDGE). The framework models temporal dependencies and embeds data using a graph representation learning method. A machine learning algorithm is then employed to classify and predict the embedded data to detect if it is secure or not. Experimental results show that this method performs well in data security detection, surpassing several dynamic graph embedding methods by 5% in terms of AUC metrics. Furthermore, the proposed framework outperforms other dynamic baseline methods in the node classification task of large-scale graphs containing 4321477 temporal information edges, resulting in a 10% improvement in the F1 score metric. The framework is also robust and scalable for application in various data security domains. This work is important for promoting the use of continuous-time graph embedding framework in the field of data security.

A Combined Deep Learning and Prior Knowledge Constraint Approach for Large-Scale Forest Disturbance Detection Using Time Series Remote Sensing Data

The scale and severity of forest disturbances across the globe are increasing due to climate change and human activities. Remote sensing analysis using time series data is a powerful approach for detecting large-scale forest disturbances and describing detailed forest dynamics. Various large-scale forest disturbance detection algorithms have been proposed, but most of them are only suitable for detecting high-magnitude forest disturbances (e.g., fire, harvest). Conversely, more continuous, subtle, and gradual lower-magnitude forest disturbances (e.g., thinning, pests, and diseases) have been subject to less focus. Deep learning (DL) can distinguish subtle differences in information within time series data, offering new opportunities to capture forest disturbances in a complete and detailed way. This study proposes an approach for analyzing forest dynamics across large areas and long time periods by combining DL time series classification and prior knowledge constraint. The approach consists of two stages: (1) an improved self-attention model used for time series classification to identify sequences with forest disturbance characteristics; (2) developed skip-disturbance recovery index (S-DRI) characterizing the temporal context, using prior knowledge constraint to identify forest disturbance years in time series with disturbance characteristics. In this study, the year of forest disturbances in five study areas located in the United States, Canada, and Poland from 2001 to 2020 was mapped. A total of 3082 manually interpreted test data with different disturbance causal agents (such as fire, harvest, conversion, hurricane, and pests) were sampled from five research areas for validation. Our approach was also evaluated against two forest disturbance benchmark datasets derived from LandTrendr and the Global Forest Change (GFC) dataset. The results demonstrate that our approach achieved an overall accuracy of 87.8%, surpassing the accuracy of LandTrendr (84.6%) and the Global Forest Change dataset (81.4%). Furthermore, our approach demonstrated lower omission rates (ranging from 10.0% to 67.4%) in detecting subtle to severe causal agents of forest disturbance, in comparison to LandTrendr (with a range of 18.0% to 81.6%) and GFC (with a range of 15.0% to 88.8%). This study, which involved mapping large-scale and long-term forest disturbance in multiple regions, revealed that our approach can be applied to new areas without a requirement for complex parameter adjustments. These results demonstrate the potential of our approach in generating comprehensive and detailed forest disturbance data, thus providing a new and effective method in this domain.

All-Optical Multiplexed Meta-Differentiator for Tri-Mode Surface Morphology Observation.

Current optical differentiators are generally limited to realizing a single differential function once fabricated. Herein, a minimalist strategy in designing multiplexed differentiators (1st - and 2nd -order differentiations), implemented with a Malus metasurface consisting of single-sized nanostructures is proposed, thus improving the functionality of optical computing devices without the cost of complex design and nanofabrication. It is found that the proposed meta-differentiator exhibits excellent differential-computation performance and can be used for simultaneous outline detection and edge positioning of objects, corresponding to the functions of the 1st - and 2nd -order differentiations respectively. Experiments with biological specimens showcase that boundaries of biological tissues can not only be identified, but also the edge information for realizing high-precision edge positioning is highlighted. The study provides a paradigm in designing all-optical multiplexed computing meta-devices, and initiates tri-mode surface morphology observation by combining meta-differentiator with optical microscopes, which can find their applications in advanced biological imaging, large-scale defect detection, and high-speed pattern recognition, etc.

Analysis and simulation of a stochastic COVID-19 model with large-scale nucleic acid detection and isolation measures: A case study of the outbreak in Urumqi, China in August 2022.

In this paper, a stochastic COVID-19 model with large-scale nucleic acid detection and isolation measures is proposed. Firstly, the existence and uniqueness of the global positive solution is obtained. Secondly, threshold criteria for the stochastic extinction and persistence in the mean with probability one are established. Moreover, a sufficient condition for the existence of unique ergodic stationary distribution for any positive solution is also established. Finally, numerical simulations are carried out in combination with real COVID-19 data from Urumqi, China and the theoretical results are verified.

YOLOv5s-M: A deep learning network model for road pavement damage detection from urban street-view imagery

Road pavement damage affects driving comfort markedly, threatens driving safety, and may even cause traffic accidents. The traffic management department conventionally captures pavement damage information mainly using manual and vehicle-mounted equipment, which is not conducive to the detection of large-scale road pavement distress. Street-view images can provide full-view images of urban roads where the data is updated regularly by navigation map service companies, making it possible to rapidly detect pavement damage in urban areas. This paper presents a new pavement damage detection approach that is built upon an improved YOLOv5 network and street-view images. The proposed model can deal with a large-scale detection layer to improve the detection precision of large distress targets, achieving thus both cross-layer and cross-scale feature fusion by using the Generalized Feature Pyramid Network (Generalized-FPN) structure. The improved network also involves a diagonal Intersection over Union loss for regression calculation of the boundary box and builds the decoupled Head structure to achieve the decoupling detection of prediction and regression. As a result, the fusion of the weak feature information in feature layers is enhanced at different spatial scales, a more suitable method is achieved for pavement damage detection in the complex context of multi-scale street-view images, and the accuracy of the modified network is much improved in the detection of pavement distress from street-view imagery. Furthermore, We created a large image sample set for model training and testing, and a total of 156,304 street-view images, obtained from Fengtai District, Beijing, China was used for demonstrating the usefulness of the proposed network. The findings indicated that the proposed approach could effectively achieve pavement damage detection of urban roads from street-view images, with a precision average of 79.8% on the test samples. Moreover, the developed model was applied for pavement damage detection for all the roads in Fengtai District, Beijing, indicating that our method can offer viable damage data for road maintenance planning.

A Lightweight Radar Ship Detection Framework with Hybrid Attentions

One of the current research areas in the synthetic aperture radar (SAR) processing fields is deep learning-based ship detection in SAR imagery. Recently, ship detection in SAR images has achieved continuous breakthroughs in detection precision. However, determining how to strike a better balance between the precision and complexity of the algorithm is very meaningful for real-time object detection in real SAR application scenarios, and has attracted extensive attention from scholars. In this paper, a lightweight object detection framework for radar ship detection named multiple hybrid attentions ship detector (MHASD) with multiple hybrid attention mechanisms is proposed. It aims to reduce the complexity without loss of detection precision. First, considering that the ship features in SAR images are not inconspicuous compared with other images, a hybrid attention residual module (HARM) is developed in the deep-level layer to obtain features rapidly and effectively via the local channel attention and the parallel self-attentions. Meanwhile, it is also capable of ensuring high detection precision of the model. Second, an attention-based feature fusion scheme (AFFS) is proposed in the model neck to further heighten the features of the object. Meanwhile, AFFS constructs and develops a fresh hybrid attention feature fusion module (HAFFM) upon the local channel and spatial attentions to guarantee the applicability of the detection model. The Large-Scale SAR Ship Detection Dataset-v1.0 (LS-SSDD-v1.0) experimental results demonstrate that MHASD can balance detection speed and precision (improving average precision by 1.2% and achieving 13.7 GFLOPS). More importantly, extensive experiments on the SAR Ship Detection Dataset (SSDD) demonstrate that the proposed method is less affected by the background such as ports and rocks.

NIS2+™, an optimisation of the blood-based biomarker NIS4® technology for the detection of at-risk NASH: A prospective derivation and validation study

NIS4® is a blood-based non-invasive test designed to effectively rule in/rule out at-risk non-alcoholic steatohepatitis (NASH), defined as non-alcoholic fatty liver disease activity score ≥4 and significant fibrosis (stage ≥2), among patients with metabolic risk factors. Robustness of non-invasive test scores across characteristics of interest including age, type 2 diabetes mellitus, and sex, and optimised analytical aspects are critical for large-scale implementation in clinical practice. We developed and validated NIS2+™, an optimisation of NIS4®, specifically designed to improve score robustness. A well-balanced training cohort (n= 198) included patients from the GOLDEN-505 trial. The validation (n= 684) and test (n= 2,035) cohorts included patients from the RESOLVE-IT trial. Well-matched subgroups were created to avoid potential confounding effects during modelling and analysis of score robustness. Models were trained using logistic regressions for at-risk NASH detection and compared using Bayesian information criteria. Performance of NIS2+™ was compared with that of NIS4®, Fibrosis-4, and alanine aminotransferase using area under the receiver operating characteristic curve, and robustness was analysed through score distribution. Using the training cohort to compare all combinations of NIS4® biomarkers, NIS2 (miR-34a-5p, YKL-40) was identified as the best combination of parameters. To correct for the sex effect on miR-34a-5p (validation cohort), sex and sex ∗ miR-34a-5p parameters were added, creating NIS2+™. In the test cohort, NIS2+™ exhibited a statistically higher area under the receiver operating characteristic curve (0.813) vs. NIS4® (0.792; p= 0.0002), Fibrosis-4 (0.653; p <0.0001), and alanine aminotransferase (0.699; p <0.0001). NIS2+™ scores were not affected by age, sex, BMI, or type 2 diabetes mellitus status, providing robust clinical performances irrespective of patient characteristics. NIS2+™ constitutes a robust optimisation of NIS4® technology for the detection of at-risk NASH. The development of non-invasive tests for accurate, large-scale detection of patients with at-risk non-alcoholic steatohepatitis (NASH; defined as NASH with non-alcoholic fatty liver disease activity score ≥4 and fibrosis stage ≥2) - who are at higher risk for disease progression and for developing liver-related life-threatening outcomes - is critical for identifying this patient population in the clinical setting and improving the screening process of NASH clinical trials. We report the development and validation of NIS2+™, a diagnostic test designed as an optimisation of NIS4® technology, a blood-based panel currently used to detect at-risk NASH in patients with metabolic risk factors. NIS2+™ showed improved performance for the detection of at-risk NASH compared with NIS4® and other non-invasive liver tests that was not impacted by patients' characteristics of interest, such as age, sex, type 2 diabetes mellitus, BMI, dyslipidaemia, and hypertension. This makes NIS2+™ a robust and reliable tool for the diagnosis of at-risk NASH among patients with metabolic risk factors, and an effective candidate for large-scale implementation in clinical practice and clinical trials.

Deep Learning for Medium-Scale Agricultural Crop Detection through Aerial View Images

This research project focuses on utilizing two state-of-the-art YOLOv4-based deep learning models, for large-scale agricultural crop detection using Unmanned Aerial Vehicles (UAVs). The objective is to develop an accurate and efficient crop detection system capable of identifying chili crops, eggplant crops, and empty polybags in agricultural fields. Crops detection is important for the development of a robotic vision in maximize the productivity and efficiency in agriculture associate with the development of concept Industry 4.0. This study seek to explore the comparison between YOLOv4 and YOLOv4 tiny model in term of mean average precision (mAP), precision, recall, F1-score, detection time and memory consumption. A custom dataset with 300 images was collected and annotated into total bounding boxes of 23335 with 6969 chili tree, 15402 eggplant tree and 964 empty polybag. The dataset was separated into train, validation and test set with the ratio of 70:20:10. The dataset was trained into YOLOv4 and YOLOv4 tiny with 2000 iterations. The result has shown that the YOLOv4 has the higher mean AP of 91.49% with 244.2mb memory storage consumption while YOLOv4 tiny achieve lower mean AP of 71.83% with 22.4mb. In summary, this research has significated the implementation of deep learning models to perform large-scale agricultural crop detection and can be further develop into automation industrial 4.0 of local agricultural sector.

Improving multi-scale detection layers in the deep learning network for wheat spike detection based on interpretive analysis

BackgroundDetecting and counting wheat spikes is essential for predicting and measuring wheat yield. However, current wheat spike detection researches often directly apply the new network structure. There are few studies that can combine the prior knowledge of wheat spike size characteristics to design a suitable wheat spike detection model. It remains unclear whether the complex detection layers of the network play their intended role.ResultsThis study proposes an interpretive analysis method for quantitatively evaluating the role of three-scale detection layers in a deep learning-based wheat spike detection model. The attention scores in each detection layer of the YOLOv5 network are calculated using the Gradient-weighted Class Activation Mapping (Grad-CAM) algorithm, which compares the prior labeled wheat spike bounding boxes with the attention areas of the network. By refining the multi-scale detection layers using the attention scores, a better wheat spike detection network is obtained. The experiments on the Global Wheat Head Detection (GWHD) dataset show that the large-scale detection layer performs poorly, while the medium-scale detection layer performs best among the three-scale detection layers. Consequently, the large-scale detection layer is removed, a micro-scale detection layer is added, and the feature extraction ability in the medium-scale detection layer is enhanced. The refined model increases the detection accuracy and reduces the network complexity by decreasing the network parameters.ConclusionThe proposed interpretive analysis method to evaluate the contribution of different detection layers in the wheat spike detection network and provide a correct network improvement scheme. The findings of this study will offer a useful reference for future applications of deep network refinement in this field.

Improving the robustness of analog deep neural networks through a Bayes-optimized noise injection approach

Analog deep neural networks (DNNs) provide a promising solution, especially for deployment on resource-limited platforms, for example in mobile settings. However, the practicability of analog DNNs has been limited by their instability due to multi-factor reasons from manufacturing, thermal noise, etc. Here, we present a theoretically guaranteed noise injection approach to improve the robustness of analog DNNs without any hardware modifications or sacrifice of accuracy, which proves that within a certain range of parameter perturbations, the prediction results would not change. Experimental results demonstrate that our algorithmic framework can outperform state-of-the-art methods on tasks including image classification, object detection, and large-scale point cloud object detection in autonomous driving by a factor of 10 to 100. Together, our results may serve as a way to ensure the robustness of analog deep neural network systems, especially for safety-critical applications.

Research on detection method of Tubercle Bacilli based on the improved YOLOv5

In the field of medical imaging, the detection method of Tubercle Bacilli based on deep learning can overcome the shortcomings of traditional manual detection methods, such as large subjectivity, large workload, slow detection speed, and reduce the occurrence of false detection or missed detection under specific circumstances. However, due to the small target and complex background of Tubercle Bacilli, the detection results are still not accurate enough. In order to reduce the influence of sputum sample background on Tubercle Bacilli detection and improve the accuracy of the model for Tubercle Bacilli detection, a target detection algorithm YOLOv5-CTS based on YOLOv5 algorithm is proposed in this paper. The algorithm first integrates the CTR3 module at the bottom of Backbone of YOLOv5 network to obtain more high-quality feature information, which brings significant performance improvement to the model; then in the neck and head part, a hybrid model with the improved feature pyramid networks and the added large-scale detection layer is utilized to perform feature fusion and small target detection; finally, the SCYLLA-Intersection over Union loss function is integrated. The experimental results show that YOLOv5-CTS increases the mean average precision to 86.2% compared with the existing target detection algorithms for Tubercle Bacilli, such as Faster R-CNN, SSD and RetinaNet, etc, which shows the effectiveness of this method.

GroupPlate: Toward Multi-Category License Plate Recognition

License Plate Detection and Recognition (LPDR) is widely used in Intelligent Transportation Systems (ITS). Although there are typically multiple categories of license plates, the majority of existing research cannot be applied to multi-category plates due to that existing methods are not optimised for multi-category plate scenarios and the scarcity of large-scale multi-category plate datasets. In this paper, we propose a multi-category license plate recognition framework called GroupPlate, which consists of Group Module and Indirect Supervision Module, making full use of the implicit and explicit grouping information of license plate. In addition, the Category Decouple Module is intended to decouple the grouping information from the original features, allowing the decoder to concentrate on character features. Simultaneously, we propose a large-scale All-Category license Plate detection and recognition Dataset (ACPD) for vehicles on the Chinese mainland, which also includes annotations of plates’ categories. Considering the domain gap between synthetic data and real data, we propose a simple but effective strategy called Feature Shift to mitigate the performance degradation caused by this gap. Experiments demonstrate that GroupPlate achieves the comparable performance to the existing methods on single-category license plate dataset and outperforms our baseline on the multi-category license plates dataset. Ablation experiments demonstrate the effectiveness of the modules in GroupPlate. Extensive results demonstrate that the dataset we proposed can mitigate the problem of models trained on a single-category license plate dataset failing to recognize multi-category license plates, and that our model can generalizes well to unseen categories. The work will be available at https://github.com/YilinGao-SHU/ACPD.

Deep Learning and Cloud-Based Computation for Cervical Spine Fracture Detection System

Modern machine learning models, such as vision transformers (ViT), have been shown to outperform convolutional neural networks (CNNs) while using fewer computational resources. Although computed tomography (CT) is now the standard for imaging diagnosis of adult spine fractures, analyzing CT scans by hand is both time consuming and error prone. Deep learning (DL) techniques can offer more effective methods for detecting fractures, and with the increasing availability of ubiquitous cloud resources, implementing such systems worldwide is becoming more feasible. This study aims to evaluate the effectiveness of ViT for detecting cervical spine fractures. Data gathered during the research indicates that ViT models are suitable for large-scale automatic detection system implementation. The model achieved 98% accuracy and was easy to train while also being easily explainable.