Average Precision Values Research Articles

Deep Learning Model Based on You Only Look Once Algorithm for Detection and Visualization of Fracture Areas in Three-Dimensional Skeletal Images.

Utilizing "You only look once" (YOLO) v4 AI offers valuable support in fracture detection and diagnostic decision-making. The purpose of this study was to help doctors to detect and diagnose fractures more accurately and intuitively, with fewer errors. The data accepted into the backbone are diversified through CSPDarkNet-53. Feature maps are extracted using Spatial Pyramid Pooling and a Path Aggregation Network in the neck part. The head part aggregates and generates the final output. All bounding boxes by the YOLO v4 are mapped onto the 3D reconstructed bone images after being resized to match the same region as shown in the 2D CT images. The YOLO v4-based AI model was evaluated through precision-recall (PR) curves and the intersection over union (IoU). Our proposed system facilitated an intuitive display of the fractured area through a distinctive red mask overlaid on the 3D reconstructed bone images. The high average precision values (>0.60) were reported as 0.71 and 0.81 from the PR curves of the tibia and elbow, respectively. The IoU values were calculated as 0.6327 (tibia) and 0.6638 (elbow). When utilized by orthopedic surgeons in real clinical scenarios, this AI-powered 3D diagnosis support system could enable a quick and accurate trauma diagnosis.

Objects detection theory for evaluating the city environmental quality

IntroductionThe primary focus of this paper is to assess urban ecological environments by employing object detection on spatial-temporal data images within a city, in conjunction with other relevant information through data mining.MethodsFirstly, an improved YOLOv7 algorithm is applied to conduct object detection, particularly counting vehicles and pedestrians within the urban spatial-temporal data. Subsequently, the k-means superpixel segmentation algorithm is utilized to calculate vegetation coverage within the urban spatial-temporal data, allowing for the quantification of vegetation area. This approach involves the segmentation of vegetation areas based on color characteristics, providing the vegetation area’s measurements. Lastly, an ecological assessment of the current urban environment is conducted based on the gathered data on human and vehicle density, along with vegetation coverage.ResultsThe enhanced YOLOv7 algorithm employed in this study yields a one-percent improvement in mean AP (average precision) compared to the original YOLOv7 algorithm. Furthermore, the AP values for key categories of interest, namely, individuals and vehicles, have also improved in this ecological assessment.DiscussionSpecifically, the AP values for the ‘person’ and ‘pedestrian’ categories have increased by 13.9% and 9.3%, respectively, while ‘car’ and ‘van’ categories have seen AP improvements of 6.7% and 4.9%. The enhanced YOLOv7 algorithm contributes to more accurate data collection regarding individuals and vehicles in subsequent research. In the conclusion of this paper, we further validate the reliability of the urban environmental assessment results by employing the Recall-Precision curve.

Content-based filtering using cosine similarity algorithm for alternative selection on training programs

The large selection of training programs provided by the Ministry of Manpower of the Republic of Indonesia makes it difficult for prospective trainees to choose a training program that suits their interests and needs. The purpose of this research is to support the selection process so that an appropriate method is needed to recommend the selection of training programs that match the interests and needs of users. One of the selection methods that can be used is the Content-Based Filtering method with similarity measurement using Cosine Similarity. The content-based filtering method is a content-based filtering method, which recommends training programs based on the suitability between the description of the training program and the interests of prospective trainees using the cosine similarity distance measurement. The test results using the Content-Based Filtering method were able to achieve an average precision value of 88%, indicating the ability of the system to provide training program recommendations that are very relevant and in accordance with the interests and needs of the trainees.

Aluminum Surface Defect Detection Algorithm Based on Improved YOLOv5

AbstractIndustrial defect detection is an important aspect of object detection. Aluminum is an indispensable material in the industrial field, but the complexity of surface defects on aluminum makes detection challenging. Therefore, the paper proposes the YOLOv5‐ESP algorithm based on the YOLOv5 algorithm. First, the problem of poor signal quality and small data samples is addressed through data enhancement. Second, the YOLOv5‐ESP incorporates the Efficient Channel Attention‐C3 (ECA‐C3) module in the Backbone structure to enhance attention toward defect regions. The Spatial Pooling Pyramid Cross‐Stage Partial Convolution (SPPCSPC) module is introduced to extract multi‐scale features of defects. The Poly‐scale Convolution (PSConv) is applied at the end of the Neck structure to resolve the problem of imprecise localization resulting from significant differences in the scale of defect features. Soft Non‐Maximum Suppression (Soft‐NMS) is utilized in the Head structure to optimize the prediction boxes. Experimental results show that the proposed YOLOv5‐ESP achieves a mean Average Precision (mAP) value of 94%, outperforming YOLOv5 and other classical algorithms. Furthermore, it maintains an average recognition time of 0.019 s per image, which meets the requirements of accuracy and real‐time performance in the industry.

Soft X-ray image recognition and classification of maize seed cracks based on image enhancement and optimized YOLOv8 model

The current investigation on image recognition and internal crack detection of maize seeds primarily relies on visible light imaging. However, due to the low transmissivity of plant cells, even with image enhancement measures, the clarity of internal cracks in the images and the subsequent feature extraction process can be a trade-off. Soft X-rays possess exceptional penetration capability and offer better safety and convenience compared to hard X-rays, making them highly suitable for visualizing internal structures within plant tissues like maize seeds. In this paper, a non-invasive Imaging Technique for Image Enhancement is proposed, combining wavelet thresholding denoising, image standardization, bilateral filtering, and laplacian sharpening. This method is based on soft X-rays and successfully achieves image recognition of cracks present inside Zhengdan 958 maize seeds using an optimized YOLOv8 model. It effectively addresses challenges related to the limited light transmission of maize seeds, difficulty in crack localization, and algorithm generalization issues. The optimized YOLOv8 model demonstrates an average precision (AP) value that is 3.1% higher than that of the original model. Furthermore, by applying image enhancement, the AP value increases by 1.8%. The proposed method exhibits an average recognition accuracy of 99.66% for intact or broken seeds, an average precision of 99.87%, an average recognition recall of 99.48%, and an average single-frame image detection time of 7.49 ms in the single seed detection experiment.

Lightweight DB-YOLO Facemask Intelligent Detection and Android Application Based on Bidirectional Weighted Feature Fusion

Conventional facemask detection algorithms face challenges of insufficient accuracy, large model size, and slow computation speed, limiting their deployment in real-world scenarios, especially on edge devices. Aiming at addressing these issues, we proposed a DB-YOLO facemask intelligent detection algorithm, which is a lightweight solution that leverages bidirectional weighted feature fusion. Our method is built on the YOLOv5 algorithm model, replacing the original YOLOv5 backbone network with the lightweight ShuffleNetv2 to reduce parameters and computational requirements. Additionally, we integrated BiFPN as the feature fusion layer, enhancing the model’s detection capability for objects of various scales. Furthermore, we employed a CARAFE lightweight upsampling factor to improve the model’s perception of details and small-sized objects and the EIOU loss function to expedite model convergence. We validated the effectiveness of our proposed method through experiments conducted on the Pascal VOC2007+2012 and Face_Mask datasets. Our experimental results demonstrate that the DB-YOLO model boasts a compact size of approximately 1.92 M. It achieves average precision values of 70.1% and 93.5% on the Pascal VOC2007+2012 and Face_Mask datasets, respectively, showcasing a 2.3% improvement in average precision compared to the original YOLOv5s. Furthermore, the model’s size is reduced by 85.8%. We also successfully deployed the model on Android devices using the NCNN framework, achieving a detection speed of up to 33 frames per second. Compared to lightweight algorithm models like YOLOv5n, YOLOv4-Tiny, and YOLOv3-Tiny, DB-YOLO not only reduces the model’s size but also effectively improves detection accuracy, exhibiting excellent practicality and promotional value on edge devices.

Rekognisi Huruf Tulisan Tangan Menggunakan Convolutional Neural Network

The development of technology in the field of computer vision in recent years with the application of Technological developments in the field of computer vision in recent years with the application of convolutional neural networks have shown sophisticated performance with a high level of accuracy, such as object detection. The problem in the world of computer vision that has been looking for a solution for a long time is object classification in general images. How to duplicate the human ability to understand images, so that computers can recognize objects in images like humans. Therefore, the need for deep learning is one branch of machine learning where the algorithm used is inspired by the workings of the human brain. Some people may be more familiar with Convolution Neural Network. CNN is used to recognize and classify patterns in handwriting. The network assumes that the input used is an image. The network has a special layer called the convolution layer. In this layer, the images are inserted according to the predefined filters. In this study, various combinations of CNN architectural designs were carried out such as the number of convolution layers, stride size, number of epochs, type of kernel size optimizer. The research data comes from the National Institute of Standards and Technology (NIST) database, then the data is divided into three, namely 60% training data, 20% validation and 20% testing. The results of this experiment produce a very good accuracy value using 2 convolution layers, 50 epochs, with Adam optimizer producing an accuracy value of 99.5% when testing the model. Then evaluate the model using the confusion matrix, assigning a high value with an average value of 100% accuracy, while for the average value of precision with a value of 100%, for an average recall value of 100%, and finally an average value of f1 score of 100%.

Detection of safety helmet and mask wearing using improved YOLOv5s

With the advancement of society, ensuring the safety of personnel involved in municipal construction projects, particularly in the context of pandemic control measures, has become a matter of utmost importance. This paper introduces a security measure for municipal engineering, combining deep learning with object detection technology. It proposes a lightweight artificial intelligence (AI) detection method capable of simultaneously identifying individuals wearing masks and safety helmets. The method primarily incorporates the ShuffleNetv2 feature extraction mechanism within the framework of the YOLOv5s network to reduce computational overhead. Additionally, it employs the ECA attention mechanism and optimized loss functions to generate feature maps with more comprehensive information, thereby enhancing the precision of target detection. Experimental results indicate that this algorithm improves the mean average precision (mAP) value by 4.3%. Furthermore, it reduces parameter and computational loads by 54.8% and 53.8%, respectively, effectively striking a balance between lightweight operation and precision. This study serves as a valuable reference for research pertaining to lightweight target detection in the realm of municipal construction safety measures.

AN IMPROVED MASK R-CNN: EXTRACTION OF DOOR AND WINDOW INSTANCES ON VILLAGE BUILDING FAÇADE IMAGES

Abstract. Rapid access to the basic structure of village buildings is conducive to the investigation of the load-bearing bodies of village houses and provides data support for disaster assessment and post-disaster rescue and reconstruction. The development of computer vision technology provides new ideas and tools for identifying and extracting basic structures of housing buildings. Considering that the original Mask R-CNN ignores the spatial association and relationship of door and window elements, an advanced deep learning model based on Mask R-CNN network is proposed in this paper to detect and segment the door and window structure from the façade images. The improved network architectures integrate the attention mechanism with the original network, containing an improved Coordinate Attention(CA) module and a relationship module-based head network. The experimental results show that the Average Precision(AP) value of the backbone combined with the improved CA module is increased by 0.7% and 0.7% on regression and segmentation tasks respectively, compared with the original Mask R-CNN network. In the head network based on the relationship module, the calculation strategy of the relational module proposed in this paper increases the AP values of detection and segmentation from 76.7% and 77.7% to 80.6% and 80.0%, respectively.

Classification Analysis of Industrial Buildings Using Orthophoto and NDSM Data with Deep Learning Approach (Case Study: Kali Rungkut Village, Surabaya)

Information of industrial estates is important because it may increase the economy level, industrial goods production, and export activities. In addition, information of industrial areas needs to be identified so that the area does not interfere with agricultural productivity, natural resources, and cultural heritage. To obtain this information, extraction and classification of building footprints using orthophoto data with a deep learning approach is carried out. However, this has the challenge that the condition of the building is highly diverse in both shape and size, so it requires additional data such as height data form (NDSM) to facilitate its identification. The Mask Region-based Convolutional Neural Network (Mask R-CNN) method used for extraction produces 88.49% precision accuracy; 66.82% completeness (recall); and 76.15% F1-score. The classification model performed with the Deep Neural Network (DNN) method, produced excellent accuracy with average values of precision, recall, and F1-score of 0.94; 0.90; and 0.92, respectively.

A deep learning approach for object detection of rockfish in challenging underwater environments

IntroductionPreserving the marine ecological environment and safeguarding marine species is a global priority. However, human overfishing has led to a drastic decline in fish species with longer growth cycles, disrupting the equilibrium of the marine ecosystem. To address this issue, researchers are turning to deep learning techniques and state-of-the-art underwater devices, such as underwater robots, to explore the aquatic environment and monitor the activities of endangered populations. This approach has emerged as a focal point of recent research in protecting the marine ecological environment. This study employs a deep learning-based object detection algorithm to identify fish species in complex underwater environments.MethodsThe algorithm is built upon the You Only Look Once version 7(YOLOv7) algorithm, with the addition of the attention mechanism Convolutional Block Attention Module (CBAM) in the network’s backbone. CBAM enhances the feature maps through the fusion of spatial attention and channel attention, ultimately improving the robustness and accuracy of the model’s inference by replacing the original loss function CIoU with SCYLLAIntersection over Union(SIoU). In this paper, the rockfish pictures in the dataset Label Fishes in the Wild published by the National Marine Fisheries Service are selected, and the underwater image enhancement model (UWCNN) is introduced to process the pictures.ResultThe experimental results show that the mean average precision (mAP) value of the improved model on the test set is 94.4%, which is 3.5% higher than the original YOLOv7 model, and the precision and recall rate are 99.1% and 99%, respectively. The detection performance of the algorithm in the field of complex underwater environment is improved.DiscussionThe underwater fish detection scheme proposed in this study holds significant practical value and significance in promoting the conservation of marine ecosystems and the protection of fish species.

A SPH-YOLOv5x-Based Automatic System for Intra-Row Weed Control in Lettuce

Weeds have a serious impact on lettuce cultivation. Weeding is an efficient way to increase lettuce yields. Due to the increasing costs of labor and the harm of herbicides to the environment, there is an increasing need to develop a mechanical weeding robot to remove weeds. Accurate weed recognition and crop localization are prerequisites for automatic weeding in precision agriculture. In this study, an intra-row weeding system is developed based on a vision system and open/close weeding knives. This vision system combines the improved you only look once v5 (YOLOv5) identification model and the lettuce–weed localization method. Compared with models including YOLOv5s, YOLOv5m, YOLOv5l, YOLOv5n, and YOLOv5x, the optimized SPH-YOLOv5x model exhibited the best performance in identifying, with precision, recall, F1-score, and mean average precision (mAP) value of 95%, 93.32%, 94.1% and 96%, respectively. The proposed weed control system successfully removed the intra-row weeds with 80.25% accuracy at 3.28 km/h. This study demonstrates the robustness and efficacy of the automatic system for intra-row weed control in lettuce.

Embedded Field Stalk Detection Algorithm for Digging–Pulling Cassava Harvester Intelligent Clamping and Pulling Device

Cassava (Manihot esculenta Crantz) is a major tuber crop worldwide, but its mechanized harvesting is inefficient. The digging–pulling cassava harvester is the primary development direction of the cassava harvester. However, the harvester clamping–pulling mechanism cannot automatically adjust its position relative to the stalks in forward movement, which results in clamping stalks with a large off-center distance difficulty, causing large harvest losses. Thus, solving the device’s clamping location problem is the key to loss reduction in the harvester. To this end, this paper proposes a real-time detection method for field stalks based on YOLOv4. First, K-means clustering is applied to improve the consistency of cassava stalk detection boxes. Next, the improved YOLOv4 network’s backbone is replaced with MobileNetV2 + CA, resulting in the KMC-YOLO network. Then, the proposed model’s validity is demonstrated using ablation studies and comparison tests. Finally, the improved network is embedded into the NVIDIA Jetson AGX Xavier, and the model is accelerated using TensorRT, before conducting field trials. The results indicate that the KMC-YOLO achieves average precision (AP) values of 98.2%, with detection speeds of 33.6 fps. The model size is reduced by 53.08% compared with the original YOLOv4 model. The detection speed after TensorRT acceleration is 39.3 fps, which is 83.64% faster than before acceleration. Field experiments show that the embedded model detects more than 95% of the time at all three harvest illumination levels. This research contributes significantly to the development of cassava harvesters with intelligent harvesting operations.

IMPLEMENTATION OF INTERNET OF THINGS (IoT) BASED ON GOOGLE SHEETS FOR WATER QUALITY MONITORING SYSTEM

Temperature, pH, and TDS are important indicators of water quality that affect various aspects. This research aims to develop a water quality monitoring system by implementing Internet of Things (IoT) technology based on Google Sheets. Based on the tests for each sensor - temperature, pH, and TDS - the results show that the system has an excellent level of measurement precision, with an average precision value of above 95%. The findings also indicate that Google Sheets can be utilized in IoT technology and can operate on more than one distinct computer device at the same time. The use of Google Sheets in IoT technology is highly effective, flexible, and user-friendly. The measured data can be automatically stored and can be visualized effectively.".

Automatic Characterization of Boulders on Planetary Surfaces From High‐Resolution Satellite Images

AbstractBoulders form from a variety of geological processes, which their size, shape, and orientation may help us better understand. Furthermore, they represent potential hazards to spacecraft landing that need to be characterized. However, mapping individual boulders across vast areas is extremely labor‐intensive, often limiting the extent over which they are characterized and the statistical robustness of obtained boulder morphometrics. To automate boulder characterization, we use an instance segmentation neural network, Mask R‐CNN, to detect and outline boulders in high‐resolution satellite and aerial images. Our neural network, BoulderNet, was trained from a data set of >33,000 boulders in >750 image tiles from Earth, the Moon, and Mars. BoulderNet not only correctly detects the majority of boulders in images but also identifies the outline of boulders with high fidelity, achieving average precision and recall values of 72% and 64% relative to manually digitized boulders from the test data set, when only detections with intersection‐over‐union ratios >50% are considered valid. These values are similar to those obtained from human mappers. On Earth, equivalent boulder diameters, aspect ratios, and orientations extracted from predictions were benchmarked against ground measurements and yield values within ±15%, ±0.20, and ±20° of their ground‐truth values, respectively. BoulderNet achieves better boulder detection and characterization performance relative to existing methods, providing a versatile open‐source tool to characterize entire boulder fields on planetary surfaces.

Transparent Component Defect Detection Method Based on Improved YOLOv7 Algorithm

Transparent components such as glass and fiber-reinforced plastics are widely used in engineering practice, which are prone to generate defects, and change its surface and internal structure, and cause great risks to the performance and stability of products. To solve the above problems, firstly, we studied the defect detection of transparent components, proposed an improved YOLOv7 (You Only Look Once V7) algorithm, replaced the Loss function CIoU (Complete-Intersection over Union) of the network model with Wise-IoU (Wise-Integration Over Union), and raised its convergence performance. Secondly, Global Attention Mechanism (GAM) is embedded in the backbone, and a dynamic target head frame is used in the output layer to generate the standard head frame and the attention function, improving the network’s attention to micro defects and ensuring the detection accuracy of micro defects. Thirdly, an intelligent defect detection platform was designed by combining mechanical engineering, visual perception, information processing and other technologies, and 150 rounds of comparative ablation experiments were conducted on typical transparent components. The improved algorithm has raised 2.6% in Mean Average Precision (MAP) value compared to the original algorithm. The improved model has better detection performance for micro defects and higher recognition accuracy. It can effectively screen out the location and category of defects, and eliminate defective components, which is consistent with the actual engineering situation. It satisfies the actual needs of product quality testing in the production process and provides reference experience for the industrial use of defect detection methods.

Real-Time Detection and Identification of Suspects in Forensic Imagery Using Advanced YOLOv8 Object Recognition Models

Rapid advancements in artificial intelligence, machine learning, deep learning, coupled with easy access to high-capacity processing hardware, expansive organized datasets, and the evolution of artificial intelligence algorithms, have extensively influenced numerous fields. Digital Forensics is one such discipline where the application of artificial intelligence has been significantly amplified in recent years. The analysis of extensive image and video files derived from forensic evidence presents challenges in terms of time efficiency and accuracy. To surmount these challenges, artificial intelligence models can be employed to perform identification and classification processes on these data, thus expediting the resolution of forensic cases with enhanced precision. In the current study, state-of-the-art pre-trained YOLOv8 object recognition models - nano, small, medium, large, and extra-large - were utilized. These models were trained on the Wider-Face dataset with the objective of identifying suspects from images and videos sourced from digital materials in the field of digital forensics. The models achieved mean Average Precision (mAP) values of 97.513%, 98.569%, 98.763%, 98.775%, and 99.032% respectively. The YOLOv8 architecture demonstrated superior performance, outperforming the YOLOv5 architecture by a margin of 7.1% to 8.8%. To aid digital forensic experts in the detection and identification of suspicious individuals, a desktop application capable of real-time image analysis was developed.

Application of deep learning technology for temporal analysis of videofluoroscopic swallowing studies

Temporal parameters during swallowing are analyzed for objective and quantitative evaluation of videofluoroscopic swallowing studies (VFSS). Manual analysis by clinicians is time-consuming, complicated and prone to human error during interpretation; therefore, automated analysis using deep learning has been attempted. We aimed to develop a model for the automatic measurement of various temporal parameters of swallowing using deep learning. Overall, 547 VFSS video clips were included. Seven temporal parameters were manually measured by two physiatrists as ground-truth data: oral phase duration, pharyngeal delay time, pharyngeal response time, pharyngeal transit time, laryngeal vestibule closure reaction time, laryngeal vestibule closure duration, and upper esophageal sphincter opening duration. ResNet3D was selected as the base model for the deep learning of temporal parameters. The performances of ResNet3D variants were compared with those of the VGG and I3D models used previously. The average accuracy of the proposed ResNet3D variants was from 0.901 to 0.981. The F1 scores and average precision were 0.794 to 0.941 and 0.714 to 0.899, respectively. Compared to the VGG and I3D models, our model achieved the best results in terms of accuracy, F1 score, and average precision values. Through the clinical application of this automatic model, temporal analysis of VFSS will be easier and more accurate.

Research on the Strawberry Recognition Algorithm Based on Deep Learning

In view of the time-consuming and laborious manual picking and sorting of strawberries, the direct impact of image recognition accuracy on automatic picking and the rapid development of deep learning(DL), a Faster Regions with Convolutional Neural Network features (R-CNN) strawberry recognition method that combines Mixup data augmentation, a ResNet(Residual Network)50 backbone feature extraction network and a Soft-NMS (Non-Maximum Suppression) algorithm, named the MRS Faster R-CNN, is proposed. In this paper, the transfer learning backbone feature extraction network VGG (Visual Geometry Group) 16 and ResNet50 are compared, and the superior ResNet50 is selected as the backbone network of MRS Faster R-CNN. The data augmentation method of Mixup image fusion is used to improve the learning and generalization ability of the model. The redundant bboxes (bounding boxes) are removed through Soft-NMS to obtain the best region proposal. The freezing phase is added to the training process, effectively reducing the occupation of video memory and shortening the training time. After experimental verification, the optimized model improved the AP (Average Precision) values of mature and immature strawberries by 0.26% and 5.34%, respectively, and the P(Precision) values by 0.81% and 6.34%, respectively, compared to the original model (R Faster R-CNN). Therefore, the MRS Faster R-CNN model proposed in this paper has great potential in the field of strawberry recognition and maturity classification and improves the recognition rate of small fruit and overlapping occluded fruit, thus providing an excellent solution for mechanized picking and sorting.

Continuous identification of the tea shoot tip and accurate positioning of picking points for a harvesting from standard plantations.

To address the current problems of large positioning error, low picking efficiency, and high cost of tea shoot picking, a continuous and precise harvesting scheme for tea shoots based on a two-dimensional (2D) perspective is designed in this study. A high-speed harvesting method for tea shoots in a standardized tea plantation assembly line type was proposed. First, a 2D view recognition model of tea shoot tips in a multi-disturbance environment was constructed, and accurate picking point coordinates were determined by combining a skeleton algorithm and curve growth. To avoid the losses of recognition accuracy caused by the mistaken clamping of blades and vibrations during harvester operations, accurate control of the harvester was realized by combining path planning and the S-curve speed control function. The recognition accuracy for the verification set of the recognition model was 99.9%, and the mean average precision (0.5:0.95) value was 0.97. The test results show that the error between the actual picking point position and the position determined by the model was within ± 3 mm, and the picking success rate was 83.6%. Therefore, we can realize fast and accurate picking of tea shoots and lay the foundation for continuous tea picking in the future by simplifying the identification and picking process.