Average Precision Values Research Articles

基于轻量化的YOLOv8小麦籽粒自动计数研究

In order to accurately and quickly achieve wheat grain detection and counting, and to efficiently evaluate wheat quality and yield, a lightweight YOLOv8 algorithm is proposed to automatically count wheat grains in different scenarios. Firstly, wheat grain images are collected under three scenarios: no adhesion, slight adhesion, and severe adhesion, to create a dataset. Then, the neck network of YOLOv8 is modified to a bidirectional weighted fusion BiFPN to establish the wheat grain detection model. Finally, the results of wheat grain counting are statistically analyzed. Experimental results show that after lightweight improvement of YOLOv8 with BiFPN, the mAP (mean Average Precision) value of wheat grain detection is 94.7%, with a reduction of 12.3% in GFLOPs. The improved YOLOv8 model now requires only 9.34ms for inference and occupies just 4.0MB of memory. Compared with other models, the proposed model in this paper performs the best in terms detection accuracy and speed comprehensively, better meeting the real-time counting requirements of wheat grains.

Extraction of grassland irrigation information in arid regions based on multi-source remote sensing data

Irrigation is a vital measure for maintaining grassland productivity in arid and semi-arid regions. Grasslands typically have characteristics such as unclear boundaries, complex vegetation types, and relatively small irrigation amounts, making it challenging to extract irrigation information. Currently, research on extracting grassland irrigation information is scarce. This study proposes a method for extracting grassland irrigation information using high spatiotemporal resolution (30 m, 1 day) downscaled surface soil moisture data, combined with Landsat 8/9 and Sentinel 1/2 data. This method was applied to extract the irrigation area, timing, and frequency of grasslands in the Delingha Piedmont, northwestern China. The results showed that the overall classification accuracy of irrigated grassland was 93.43 %, and the kappa coefficient was 0.91, indicating high extraction accuracy. The average values of recall, precision, and F-score for irrigation timing and frequency were 82.54 %, 72.25 %, and 77.03 %, respectively, with most irrigation events accurately identified, indicating commendable overall efficacy. The combined use of multi-source remote sensing data is crucial for the extraction of grassland irrigation information. Among these, The high spatiotemporal resolution downscaled surface soil moisture data, by providing detailed spatiotemporal surface soil moisture dynamics, demonstrate a potent capacity for capturing irrigation events, thus effectively enhancing the accuracy of grassland irrigation data extraction.

Towards accurate and efficient diagnoses in nephropathology: An AI-based approach for assessing kidney transplant rejection

The Banff classification is useful for diagnosing renal transplant rejection. However, it has limitations due to subjectivity and varying concordance in physicians' assessments. Artificial intelligence (AI) can help standardize research, increase objectivity and accurately quantify morphological characteristics, improving reproducibility in clinical practice. This study aims to develop an AI-based solutions for diagnosing acute kidney transplant rejection by introducing automated evaluation of prognostic morphological patterns. The proposed approach aims to help accurately distinguish borderline changes from rejection. We trained a deep-learning model utilizing a fine-tuned Mask R-CNN architecture which achieved a mean Average Precision value of 0.74 for the segmentation of renal tissue structures. A strong positive nonlinear correlation was found between the measured infiltration areas and fibrosis, indicating the model's potential for assessing these parameters in kidney biopsies. The ROC analysis showed a high predictive ability for distinguishing between ci and i scores based on infiltration area and fibrosis area measurements. The AI model demonstrated high precision in predicting clinical scores which makes it a promising AI assisting tool for pathologists. The application of AI in nephropathology has a potential for advancements, including automated morphometric evaluation, 3D histological models and faster processing to enhance diagnostic accuracy and efficiency.

A Machine Learning Prediction Model for Myelitis and Multiple Sclerosis Based on Fourier Transform Features from MRI Images

Myelitis is a neurodegenerative disease positioned in the spinal cord, with multiple sclerosis (MS) being a common subtype. Radiological indicators enable the diagnosis of these diseases. This study proposes a classification framework to detect myelitis, MS, and healthy control (HC) groups using magnetic resonance imaging (MRI) images. The feature extraction step involves applying the fast Fourier transform (FFT) to MRI images. FFT is important because it converts spatial data into the frequency domain, making it easier to identify patterns and abnormalities that indicate these diseases. Then, statistical features (mean, minimum, maximum, standard deviation, skewness, kurtosis, and total energy) are extracted from this frequency information. These features are then used to train support vector machine (SVM), k-nearest neighbor (KNN), and decision tree algorithms. In multi-class classification (myelitis vs. MS vs. HC), the proposed method achieves a classification accuracy of 99.31% with SVM, with average precision, recall, and F1-score values of 99.27%, 99.21%, and 99.24%, respectively, indicating effective classification across all classes. In the binary class classification (HC vs. MS, MS vs. myelitis, HC vs. myelitis), the SVM achieves an outstanding classification accuracy of 99.36%, 99.71%, and 100% respectively. This study highlights the efficiency of FFT-based feature extraction in forming detection patterns for classifying HC, MS, and myelitis classes.

Breaking Barriers in Cancer Diagnosis: Super‐Light Compact Convolution Transformer for Colon and Lung Cancer Detection

ABSTRACTAccording to the World Health Organization, lung and colon cancers are known for their high mortality rates which necessitate the diagnosis of these cancers at an early stage. However, the limited availability of data such as histopathology images used for diagnosis of these cancers, poses a significant challenge while developing computer‐aided detection system. This makes it necessary to keep a check on the number of parameters in the artificial intelligence (AI) model used for the detection of these cancers considering the limited availability of the data. In this work, a customised compact and efficient convolution transformer architecture, termed, C3‐Transformer has been proposed for the diagnosis of colon and lung cancers using histopathological images. The proposed C3‐Transformer relies on convolutional tokenisation and sequence pooling approach to keep a check on the number of parameters and to combine the advantage of convolution neural network with the advantages of transformer model. The novelty of the proposed method lies in efficient classification of colon and lung cancers using the proposed C3‐Transformer architecture. The performance of the proposed method has been evaluated on the ‘LC25000’ dataset. Experimental results shows that the proposed method has been able to achieve average classification accuracy, precision and recall value of 99.30%, 0.9941 and 0.9950, in classifying the five different classes of colon and lung cancer with only 0.0316 million parameters. Thus, the present computer‐aided detection system developed using proposed C3‐Transformer can efficiently detect the colon and lung cancers using histopathology images with high detection accuracy.

Using Principal Component Analysis for Temperature Readings from YF3:Pr3+ Luminescence

The method of measuring temperature using luminescence by analyzing the emission spectra of Pr3+-doped YF3 using principal component analysis is presented. The Pr3+-doped YF3 is synthesized using a solid-state technique, and its single-phase orthorhombic crystal structure is confirmed using X-ray diffraction. The emission spectra measured within the 93–473 K temperature range displays characteristic Pr3+ f-f electronic transitions. The red emission from the 3P0,1→3H6,3F2 electronic transition mostly dominates the spectra. However, at low temperatures, the intensity of the green emissions from the 3P0,1→3H5, deep-red 3P0,1→3F4, and the deep-red emissions from the 3P0,1→3F4 transitions are considerably lower compared to the intensity of the red emissions. Temperature variations directly impact the photoluminescent spectra, causing a notable increase in the green and deep-red emissions from the 3P1 excited state. We utilized the entire spectrum as an input for principal component analysis, considering each temperature as an independent group of data. The first principal component explained 99.3% of the variance in emission spectra caused by temperature and we further used it as a reliable temperature indicator for luminescence thermometry. The approach has a maximum absolute sensitivity of around 0.012 K−1. The average accuracy and precision values are 0.7 K and 0.5 K, respectively.

Disease gene prioritization with quantum walks.

Disease gene prioritization methods assign scores to genes or proteins according to their likely relevance for a given disease based on a provided set of seed genes. This scoring can be used to find new biologically relevant genes or proteins for many diseases. Although methods based on classical random walks have proven to yield competitive results, quantum walk methods have not been explored to this end. We propose a new algorithm for disease gene prioritization based on continuous-time quantum walks using the adjacency matrix of a protein-protein interaction (PPI) network. We demonstrate the success of our proposed quantum walk method by comparing it to several well-known gene prioritization methods on three disease sets, across seven different PPI networks. In order to compare these methods, we use cross-validation and examine the mean reciprocal ranks of recall and average precision values. We further validate our method by performing an enrichment analysis of the predicted genes for coronary artery disease. The data and code for the methods can be accessed at https://github.com/markgolds/qdgp.

An Automated Diagnosis Method for Lung Cancer Target Detection and Subtype Classification-Based CT Scans.

When dealing with small targets in lung cancer detection, the YOLO V8 algorithm may encounter false positives and misses. To address this issue, this study proposes an enhanced YOLO V8 detection model. The model integrates a large separable kernel attention mechanism into the C2f module to expand the information retrieval range, strengthens the extraction of lung cancer features in the Backbone section, and achieves effective interaction between multi-scale features in the Neck section, thereby enhancing feature representation and robustness. Additionally, depth-wise convolution and Coordinate Attention mechanisms are embedded in the Fast Spatial Pyramid Pooling module to reduce feature loss and improve detection accuracy. This study introduces a Minimum Point Distance-based IOU loss to enhance correlation between predicted and ground truth bounding boxes, improving adaptability and accuracy in small target detection. Experimental validation demonstrates that the improved network outperforms other mainstream detection networks in terms of average precision values and surpasses other classification networks in terms of accuracy. These findings validate the outstanding performance of the enhanced model in the localization and recognition aspects of lung cancer auxiliary diagnosis.

Automatic Perception of Typical Abnormal Situations in Cage-Reared Ducks Using Computer Vision.

Overturning and death are common abnormalities in cage-reared ducks. To achieve timely and accurate detection, this study focused on 10-day-old cage-reared ducks, which are prone to these conditions, and established prior data on such situations. Using the original YOLOv8 as the base network, multiple GAM attention mechanisms were embedded into the feature fusion part (neck) to enhance the network's focus on the abnormal regions in images of cage-reared ducks. Additionally, the Wise-IoU loss function replaced the CIoU loss function by employing a dynamic non-monotonic focusing mechanism to balance the data samples and mitigate excessive penalties from geometric parameters in the model. The image brightness was adjusted by factors of 0.85 and 1.25, and mainstream object-detection algorithms were adopted to test and compare the generalization and performance of the proposed method. Based on six key points around the head, beak, chest, tail, left foot, and right foot of cage-reared ducks, the body structure of the abnormal ducks was refined. Accurate estimation of the overturning and dead postures was achieved using the HRNet-48. The results demonstrated that the proposed method accurately recognized these states, achieving a mean Average Precision (mAP) value of 0.924, which was 1.65% higher than that of the original YOLOv8. The method effectively addressed the recognition interference caused by lighting differences, and exhibited an excellent generalization ability and comprehensive detection performance. Furthermore, the proposed abnormal cage-reared duck pose-estimation model achieved an Object Key point Similarity (OKS) value of 0.921, with a single-frame processing time of 0.528 s, accurately detecting multiple key points of the abnormal cage-reared duck bodies and generating correct posture expressions.

Improved 3D Object Detection Based on PointPillars

Despite the recent advancements in 3D object detection, the conventional 3D point cloud object detection algorithms have been found to exhibit limited accuracy for the detection of small objects. To address the challenge of poor detection of small-scale objects, this paper adopts the PointPillars algorithm as the baseline model and proposes a two-stage 3D target detection approach. As a cutting-edge solution, point cloud processing is performed using Transformer models. Additionally, a redefined attention mechanism is introduced to further enhance the detection capabilities of the algorithm. In the first stage, the algorithm uses PointPillars as the baseline model. The central concept of this algorithm is to transform the point cloud space into equal-sized columns. During the feature extraction stage, when the features from all cylinders are transformed into pseudo-images, the proposed algorithm incorporates attention mechanisms adapted from the Squeeze-and-Excitation (SE) method to emphasize and suppress feature information. Furthermore, the 2D convolution of the traditional backbone network is replaced by dynamic convolution. Concurrently, the addition of the attention mechanism further improves the feature representation ability of the network. In the second phase, the candidate frames generated in the first phase are refined using a Transformer-based approach. The proposed algorithm applies channel weighting in the decoder to enhance channel information, leading to improved detection accuracy and reduced false detections. The encoder constructs the initial point features from the candidate frames for encoding. Meanwhile, the decoder applies channel weighting to enhance the channel information, thereby improving the detection accuracy and reducing false detections. In the KITTI dataset, the experimental results verify the effectiveness of this method in small objects detection. Experimental results show that the proposed method significantly improves the detection capability of small objects compared with the baseline PointPillars. In concrete terms, in the moderate difficulty detection category, cars, pedestrians, and cyclists average precision (AP) values increased by 5.30%, 8.1%, and 10.6%, respectively. Moreover, the proposed method surpasses existing mainstream approaches in the cyclist category.

Intelligent Inspection Method and System of Plastic Gear Surface Defects Based on Adaptive Sample Weighting Deep Learning Model.

After injection molding, plastic gears often exhibit surface defects, including those on end faces and tooth surfaces. These defects encompass a wide range of types and possess complex characteristics, which pose challenges for inspection. Current visual inspection systems for plastic gears suffer from limitations such as single-category defect inspection and low accuracy. There is an urgent industry need for a comprehensive and accurate method and system for inspecting defects on plastic gears, with improved inspection capability and higher accuracy. This paper presents an intelligent inspection algorithm network for plastic gear defects (PGD-net), which effectively captures subtle defect features at arbitrary locations on the surface compared to other models. An adaptive sample weighting method is proposed and integrated into an improved Focal-IoU loss function to address the issue of low inspection accuracy caused by imbalanced defect dataset distributions, thus enhancing the regression accuracy for difficult defect categories. CoordConv layers are incorporated into each inspection head to improve the model's generalization capability. Furthermore, a dataset of plastic gear surface defects comprising 16 types of defects is constructed, and our algorithm is trained and tested on this dataset. The PGD-net achieves a comprehensive mean average precision (mAP) value of 95.6% for the 16 defect types. Additionally, an online inspection system is developed based on the PGD-net algorithm, which can be integrated with plastic gear production lines to achieve online full inspection and automatic sorting of plastic gear defects. The entire system has been successfully applied in plastic gear production lines, conducting daily inspections of over 60,000 gears.

Automated Surface Crack Detection in Historical Constructions with Various Materials Using Deep Learning-Based YOLO Network

ABSTRACT Cultural heritage (CH) constructions involve the use of diverse masonry materials. Under natural and human influences, masonry materials can undergo various types of damages, with crack damages being most prevalent. Developing a robust model using deep learning (DL) capable of detecting cracks in various CH materials is crucial. In this study, we compared the performance of the DL method You Only Look Once (YOLO) object detection network based on images in different masonry materials (stone, brick, cob, and tile) with that in a modern material (concrete). The dataset used in the study comprised 1213 bricks, 1116 concretes, 955 cobs, 882 stones, and 208 tile images. YOLOv5 architecture, transfer learning, and object detection models were utilized for detecting cracks and compare their performance in different materials. This study represents the first comparison of this kind using an original dataset. The DL model achieved mean average precision values of 94.4%, 93.9%, 92.7%, 87.2%, 83.4%, 81.6%, and 70.3% for concrete; concrete and cob, cob; stone; stone and brick; brick; and tile, respectively. The findings of this study indicate considerable potential for the widespread use of DL techniques in identifying cracks from images across various CH materials and assisting inspection professionals in damage surveys.

Using Data Augmentation to Improve the Generalization Capability of an Object Detector on Remote-Sensed Insect Trap Images.

Traditionally, monitoring insect populations involved the use of externally placed sticky paper traps, which were periodically inspected by a human operator. To automate this process, a specialized sensing device and an accurate model for detecting and counting insect pests are essential. Despite considerable progress in insect pest detector models, their practical application is hindered by the shortage of insect trap images. To attenuate the "lack of data" issue, the literature proposes data augmentation. However, our knowledge about data augmentation is still quite limited, especially in the field of insect pest detection. The aim of this experimental study was to investigate the effect of several widely used augmentation techniques and their combinations on remote-sensed trap images with the YOLOv5 (small) object detector model. This study was carried out systematically on two different datasets starting from the single geometric and photometric transformation toward their combinations. Our results show that the model's mean average precision value (mAP50) could be increased from 0.844 to 0.992 and from 0.421 to 0.727 on the two datasets using the appropriate augmentation methods combination. In addition, this study also points out that the integration of photometric image transformations into the mosaic augmentation can be more efficient than the native combination of augmentation techniques because this approach further improved the model's mAP50 values to 0.999 and 0.756 on the two test sets, respectively.

Deep Attention Fusion Hashing (DAFH) Model for Medical Image Retrieval.

To address this, we propose a novel deep learning-based hashing model, the Deep Attention Fusion Hashing (DAFH) model, which integrates advanced attention mechanisms with medical imaging data. The DAFH model enhances retrieval performance by integrating multi-modality medical imaging data and employing attention mechanisms to optimize the feature extraction process. Utilizing multimodal medical image data from the Cancer Imaging Archive (TCIA), this study constructed and trained a deep hashing network that achieves high-precision classification of various cancer types. At hash code lengths of 16, 32, and 48 bits, the model respectively attained Mean Average Precision (MAP@10) values of 0.711, 0.754, and 0.762, highlighting the potential and advantage of the DAFH model in medical image retrieval. The DAFH model demonstrates significant improvements in the efficiency and accuracy of medical image retrieval, proving to be a valuable tool in clinical settings.

YOLOX-CS: An Automatic Search Algorithm for Low Surface Brightness Galaxies

The characteristics of Low Surface Brightness Galaxies (LSBGs) are very important for understanding the overall characteristics of galaxies. It is of great significance to search and expand the samples of Low Surface Brightness Galaxies by modern machine learning, especially deep learning algorithm. LSBGs are difficult to discern automatically and accurately with traditional methods because of their obscure features. However, deep learning does have the advantage of automatically identifying complex and effective features. To solve this problem, an algorithm named You Only Look Once version X-CS (YOLOX-CS) is proposed to search LSBG in large sample sky survey. Firstly, five classical target detection algorithms are compared through experiments and the optimal YOLOX algorithm is selected as the basic algorithm. Then, the YOLOX-CS framework is constructed by combining different attention mechanisms and different optimizers. The data set uses images from the Sloan Digital Sky Survey (SDSS), labelled from LSBG in the α.40-SDSS DR7 (the cross coverage area of 40% HI Arecibo Legacy Fast ALFA Survey and SDSS Data Release7) survey. Due to the small number of samples in this data set, Deep Convolutional Generative Adversarial Networks (DCGAN) model is used to expand the experimental test data. After comparing with a series of target detection algorithms, YOLOX-CS has a good test result in searching LSBG recall rate and Average Precision (AP) value in two data sets before and after expansion. The recall rate and AP value in the test set without expansion data set reach 97.75% and 97.83%, respectively. In the expanded data set of DCGAN model, under the same test set, the recall rate reaches 99.10% and the AP value reaches 98.94%, which proves that the algorithm has excellent performance in LSBG search. Finally, the algorithm is applied to SDSS photometric data, and 765 LSBG candidates are obtained.

Estimation of Picea Schrenkiana Canopy Density at Sub-Compartment Scale by Integration of Optical and Radar Satellite Images

This study proposes a novel approach to estimate canopy density in Picea Schrenkiana var. Tianschanica forest sub-compartments by integrating optical and radar satellite data. This effort is aimed at enhancing methodologies for forest resource surveys and monitoring, particularly vital for the sustainable development of semi-arid mountainous areas with fragile ecological environments. The study area is the West Tianshan Mountain Nature Reserve in Xinjiang, which is characterized by its unique dominant tree species, Picea Schrenkiana. A total of 411 characteristic factors were extracted from Gaofen-2 (GF-2) sub-meter optical satellite imagery, Gaofen-3 (GF-3) multi-polarization synthetic aperture radar satellite imagery, and digital elevation model (DEM) data. Consequently, 17 characteristic parameters were selected based on their correlation with canopy density data to construct an estimation model. Three distinct models were developed, including a multiple stepwise regression model (a linear approach), a Back Propagation (BP) neural network model (a neural network-based method), and a Cubist model (a decision tree-based technique). The results indicate that combining optical and radar image characteristics significantly enhances accuracy, with an Average Absolute Percentage Precision (AAPP) value improvement in estimation accuracy from 76.50% (with optical image) and 78.50% (with radar image) to 78.66% (with both). Of the three models, the BP neural network model achieved the highest overall accuracy (79.19%). At the sub-component scale, the BP neural network model demonstrated superior accuracy in low canopy density estimation (75.37%), whereas the Cubist model, leveraging radar image characteristics, excelled in medium density estimations (87.46%). Notably, the integrated Cubist model combining optical and radar data achieved the highest accuracy for high canopy density estimation (89.17%). This study highlights the effectiveness of integrating optical and radar data for precise canopy density assessment, contributing significantly to ecological resource monitoring methodologies and environmental assessments.

A framework for automated reconstruction of communication towers from terrestrial laser scanning point clouds

ABSTRACT Communication towers provide reliable wireless communication services and play a vital role in modern society. This paper presents an automatic 3D reconstruction framework for communication towers based on point clouds. The antenna is reconstructed by a model-driven method, which constructs an energy function including the data and smoothing terms to evaluate the model accuracy. The parameters of the antenna model are obtained by minimizing energy function with hybrid Genetic Algorithm (GA) and Simulated Annealing (SA). And the predefined model is transformed using these parameters to achieve antenna 3D reconstruction. For tower body reconstruction, the point cloud of the tower body is finely extracted with the region growing followed by piecewise fitting to obtain the model. The performance of the proposed method is evaluated on point clouds from six scenes. The average root-mean-square error (RMSE) of antenna reconstruction is 1.3 cm. The average precision, recall, and F1-score values of the tower body point extraction are 99.9%, 93.5%, and 96.6%, respectively. The average RMSE of tower body reconstruction is 1.2 cm. The experimental results indicate that the method can effectively reconstruct the antenna and tower body with point clouds, providing a feasible framework for the 3D reconstruction of communication towers.

Cross-modal learning representation using new margin combination for speech recognition task

Cross-modal retrieval aims to elucidate the fusion of information, mimic human learning, and advance the field. The main challenge in cross-modal matching is to build a shared subspace reflecting semantic proximity. Previous works fail to capture asymmetric relevance by adopting symmetric similarity computations. To overcome these shortcomings, an efficient approach called quaternion representation learning (QRL) is introduced for better cross-modal matching. Thus, a better representation of the shared semantics is offered by virtue of its richer representation capacity of the quaternionic space and its strong expressive power. Transfer learning is a crucial aspect in this context. By leveraging pre-trained models, the knowledge gained from one task or domain can be effectively transferred to another, allowing for improved performance and generalization. In this study, transfer learning is employed to enhance the cross-modal retrieval system. Specifically, a pre-trained ResNet-512 model is utilized in conjunction with the proposed total margin (TM) loss function, which combines the QRL approach with the novel adaptive mean margin (AMM) methodology. The TM loss function, coupled with the pre-trained ResNet-512 model, is evaluated on the Audio-Visual Arabic Speech Database (AVAS) and the Arabic Visual Speech Database (AVSD), along with other audio-visual datasets. Experimental results demonstrate the effectiveness of the TM loss function in consistently improving performance on both databases. The recall scores (R@k) and mean average precision (mAP) values achieved on the AVAS Database are as follows: R@1: 42.1±0.7, R@2: 70.2±0.1, R@5: 78.5±1.0, and mAP: 53.0±1.1. Similarly, on the AVSD Database, the results are R@1: 41.7±0.3, R@2: 69.2±1.1, R@5: 78.0±0.3, and mAP: 52.7±0.5. By incorporating transfer learning and the TM loss function into the cross-modal retrieval framework, this study demonstrates the potential for improving clustering efficiency and enhancing visual and speech understanding. The combination of pre-trained models and the TM loss function offers a promising avenue for advancing crossmodal matching techniques and achieving state-of-the-art performance.

Deep Learning for Skeleton-Based Human Activity Segmentation: An Autoencoder Approach

Automatic segmentation is essential for enhancing human activity recognition, especially given the limitations of publicly available datasets that often lack diversity in daily activities. This study introduces a novel segmentation method that utilizes skeleton data for a more accurate and efficient analysis of human actions. By employing an autoencoder, this method extracts representative features and reconstructs the dataset, using the discrepancies between the original and reconstructed data to establish a segmentation threshold. This innovative approach allows for the automatic segmentation of activity datasets into distinct segments. Rigorous evaluations against ground truth across three publicly available datasets demonstrate the method’s effectiveness, achieving impressive average annotation error, precision, recall, and F1-score values of 3.6, 90%, 87%, and 88%, respectively. This illustrates the robustness of the proposed method in accurately identifying change points and segmenting continuous skeleton-based activities as compared to two other state-of-the-art techniques: one based on deep learning and another using the classical time-series segmentation algorithm. Additionally, the dynamic thresholding mechanism enhances the adaptability of the segmentation process to different activity dynamics improving overall segmentation accuracy. This performance highlights the potential of the proposed method to significantly advance the field of human activity recognition by improving the accuracy and efficiency of identifying and categorizing human movements.

YOLO-Chili: An Efficient Lightweight Network Model for Localization of Pepper Picking in Complex Environments

Currently, few deep models are applied to pepper-picking detection, and existing generalized neural networks face issues such as large model parameters, prolonged training times, and low accuracy. To address these challenges, this paper proposes the YOLO-chili target detection algorithm for chili pepper detection. Initially, the classical target detection algorithm YOLOv5 serves as the benchmark model. We introduce an adaptive spatial feature pyramid structure that combines the attention mechanism and the concept of multi-scale prediction to enhance the model’s detection capabilities for occluded and small target peppers. Subsequently, we incorporate a three-channel attention mechanism module to improve the algorithm’s long-distance recognition ability and reduce interference from redundant objects. Finally, we employ a quantized pruning method to reduce model parameters and achieve lightweight processing. Applying this method to our custom chili pepper dataset, we achieve an average precision (AP) value of 93.11% for chili pepper detection, with an accuracy rate of 93.51% and a recall rate of 92.55%. The experimental results demonstrate that YOLO-chili enables accurate and real-time pepper detection in complex orchard environments.