Object Classification Accuracy Research Articles

CelloType: A Unified Model for Segmentation and Classification of Tissue Images.

Cell segmentation and classification are critical tasks in spatial omics data analysis. We introduce CelloType, an end-to-end model designed for cell segmentation and classification of biomedical microscopy images. Unlike the traditional two-stage approach of segmentation followed by classification, CelloType adopts a multi-task learning approach that connects the segmentation and classification tasks and simultaneously boost the performance of both tasks. CelloType leverages Transformer-based deep learning techniques for enhanced accuracy of object detection, segmentation, and classification. It outperforms existing segmentation methods using ground-truths from public databases. In terms of classification, CelloType outperforms a baseline model comprised of state-of-the-art methods for individual tasks. Using multiplexed tissue images, we further demonstrate the utility of CelloType for multi-scale segmentation and classification of both cellular and non-cellular elements in a tissue. The enhanced accuracy and multi-task-learning ability of CelloType facilitate automated annotation of rapidly growing spatial omics data.

A new CNN-BASED object detection system for autonomous mobile robots based on real-world vehicle datasets

Recently, autonomous mobile robots (AMRs) have begun to be used in the delivery of goods, but one of the biggest challenges faced in this field is the navigation system that guides a robot to its destination. The navigation system must be able to identify objects in the robot's path and take evasive actions to avoid them. Developing an object detection system for an AMR requires a deep learning model that is able to achieve a high level of accuracy, with fast inference times, and a model with a compact size that can be run on embedded control systems. Consequently, object recognition requires a convolutional neural network (CNN)-based model that can yield high object classification accuracy and process data quickly. This paper introduces a new CNN-based object detection system for an AMR that employs real-world vehicle datasets. First, we create original real-world datasets of images from Banda Aceh city. We then develop a new CNN-based object identification system that is capable of identifying cars, motorcycles, people, and rickshaws under morning, afternoon, and evening lighting conditions. An SSD Mobilenetv2 FPN Lite 320 × 320 architecture is employed for retraining using these real-world datasets. Quantitative and qualitative performance indicators are then applied to evaluate the CNN model. Training the pre-trained SSD Mobilenetv2 FPN Lite 320 × 320 model improves its classification and detection accuracy, as indicated by its performance results. We conclude that the proposed CNN-based object detection system has the potential for use in an AMR.

SRGAN-enhanced unsafe operation detection and classification of heavy construction machinery using cascade learning

In the inherently hazardous construction industry, where injuries are frequent, the unsafe operation of heavy construction machinery significantly contributes to the injury and accident rates. To reduce these risks, this study introduces a novel framework for detecting and classifying these unsafe operations for five types of construction machinery. Utilizing a cascade learning architecture, the approach employs a Super-Resolution Generative Adversarial Network (SRGAN), Real-Time Detection Transformers (RT-DETR), self-DIstillation with NO labels (DINOv2), and Dilated Neighborhood Attention Transformer (DiNAT) models. The study focuses on enhancing the detection and classification of unsafe operations in construction machinery through upscaling low-resolution surveillance footage and creating detailed high-resolution inputs for the RT-DETR model. This enhancement, by leveraging temporal information, significantly improves object detection and classification accuracy. The performance of the cascaded pipeline yielded an average detection and first-level classification precision of 96%, a second-level classification accuracy of 98.83%, and a third-level classification accuracy of 98.25%, among other metrics. The cascaded integration of these models presents a well-rounded solution for near-real-time surveillance in dynamic construction environments, advancing surveillance technologies and significantly contributing to safety management within the industry.

Clustering-inspired channel selection method for weakly supervised object localization

Weakly Supervised Object Localization (WSOL) aims to utilize the features learned by a classifier on the image-level labels to locate target objects. However, these existing channel selection methods for WSOL still cannot effectively select the important channels and remove the unimportant ones. To address this issue, we propose a Clustering-inspired Channel Selection method based on Class Activation Maps (CCS-CAM). Compared with the traditional methods, the advantage of CCS-CAM is that it is very simple yet effective for channel selection due to the K-means clustering based on Class Activation Maps. It can effectively ensure both object localization and classification accuracy. The effectiveness of the proposed CCS-CAM method has been demonstrated using multiple public datasets, with GT-Know Loc reaching 87.9% and 63.71% on the CUB200-2011 and ImageNet-1k respectively, which is superior to the other state-of-the-art methods.

Monitoring dynamic mangrove landscape patterns in China: Effects of natural and anthropogenic forcings during 1985–2020

Many mangrove forests located along estuaries and deltas worldwide are currently facing irreparable losses caused by natural factors and intensive anthropogenic disturbances. The limited capabilities of traditional remote sensing processing platforms tend to lead to poor image processing consistency in medium- and large-scale areas, resulting in lower ground object classification accuracy. Consequently, little information is available regarding the extent and causes of variation in large-scale mangrove forests. In this study, we used Google Earth Engine mangrove forest data from eight periods between 1985 and 2020, with total accuracy rates of 82.95–90.15%, and combined landscape metrics with a centroid transfer model to quantitatively analyze the landscape evolution of mangrove forests in China. Finally, we employed a geodetector based on optimized parameters to quantitatively determine the potential mechanisms driving mangrove landscape evolution. Overall, a slight increase followed by a sharp decrease in mangrove forest area in China during the past 40 years, with a turning point around 2000 and substantial shifts in mangrove spatial patterns in Guangdong and Guangxi Provinces, particularly in the Pearl River Estuary and Beibu Gulf. The landscape pattern of Chinese mangrove forests has become fragmented and isolated through rapid urbanization in recent decades. Mangrove forests shifted significantly (> 31 km) early in the study period, followed by smaller shifts later in the study period. The optimal spatial scale of mangrove landscape driver detection was 270 m × 270 m. The variation in mangrove forest area was caused by the interactions of multiple factors, particularly between mean annual precipitation and other potential influencing factors. Government policies guided the conservation and management of mangrove forests in China to varying degrees throughout the study period. Our findings on the dynamics of mangrove forests provide a reference for integrated coastal management, which will contribute to the ecological utilization and high-quality development of coasts during extensive urbanization along the southeastern coast of China and similar large estuary deltas worldwide.

Study of methods for searching and localizing objects in images from aircraft using convolutional neural networks

The use of unmanned and manned aerial vehicles for remote object localization and classification is very common. These methods are used in various systems, ranging from territory surveys to law enforcement. Methods of object localization and classification using neural networks require a detailed study and research of the quality of their work on data that has certain specifics, such as vehicle detection. The use of neural networks to detect certain types of objects using images obtained from aircraft can also help in the study of hard-to-reach locations. Therefore, the main subject of this paper is the localization and classification of objects in images obtained using digital cameras mounted on aircraft. The main focus is on determining the accuracy of object localization and detection using selected types of neural networks, which are the most important indicators of neural network efficiency. The speed of a neural network is also an equally important characteristic as it directly affects its ability to be used in tasks that require fast object localization, such as video surveillance or automated car control systems. The main goal of this study is to study the accuracy of object localization and classification in images obtained with the help of cameras mounted on aircraft, as well as to study the speed of neural networks and determine the effectiveness of their application in real-world conditions. The objectives of this study are to train YOLO v5, SSD, and Faster RCNNs on the VisDrone dataset and to further study them on the vehicle localization dataset. The main goal of this work is to obtain statistics on the performance of neural networks trained on the VisDrone dataset. On the basis of the obtained statistics, conclusions are drawn about the effectiveness of the considered neural networks. The conclusions are drawn by considering the speed of the model, localization (IoU), and classification (Precision, Recall) metrics. Possible directions for further development of the topic under study are presented as conclusions.

Dynamic convolution layer based optimization techniques for object classification and semantic segmentation

<p>Providing meaningful classification for each pixel in an image is a primary goal of computer vision, and the tasks of object classification and semantic segmentation are among the field’s greatest challenges. To improve object classification, this study presents a novel method that combines semantic segmentation with dynamic convolution layer-based optimization techniques. In the proposed method, a Refined Convolution Neural Network (R-CNN) is used, which uses non-extensive entropy to dynamically increase the size of its convolutional layers. The Common Objects in Context (COCO) dataset is used to assess the performance of the model. The model performs exceptionally well at different Intersections over Union (IoU) cutoffs, with average precision values of 40.1, 61.9, and 45.4, respectively, for Average Precision (AP), AP<sub>50</sub>, and AP<sub>75</sub>. These results demonstrate the model’s efficiency in discriminating between various image contents. Additionally, the model predicts an image’s outcome on average in just 0.901 s. The model has been proven to be superior through various performance evaluation parameters, showing an average mean precision of 91.78%. This study demonstrates the power of combining dynamic convolution layers with semantic segmentation to improve object classification accuracy, a key component in the development of computer vision applications.</p>

Multilevel Fine-Grained Features-Based General Framework for Object Detection.

This article proposes a practical and generalizable object detector, termed feature extraction-fusion-prediction network (FEFP-Net) for real-world application scenarios. The existing object detection methods have recently achieved excellent performance, however they still face three major challenges for real-world applications, i.e., feature similarity between classes, object size variability, and inconsistent localization and classification predictions. In order to effectively alleviate the current difficulties, the FEFP-Net with three key components is proposed, and the improved detection accuracy is proved in various applications: 1) Extraction Phase: an adaptive fine-grained feature extraction network is proposed to capture features of interest from coarse to fine details, which effectively avoids misclassification due to feature similarity; 2) Fusion Phase: a bidirectional neighbor connection network is designed to identify objects with different sizes by aggregating multilevel features and 3) Prediction Phase: in order to improve the accuracy of object localization and classification, a task specific prediction network is presented, which sufficiently exploits both the spatial and channel information of features. Compared with the State-of-the-Art methods, we achieved competitive results in the MS-COCO dataset. Further, we demonstrated the performance of FEFP-Net in different application fields, such as medical imaging, industry, agriculture, transportation, and remote sensing. These comprehensive experiments indicate that FEFP-Net has satisfactory accuracy and generalizability as a basic object detector.

Improving the accuracy of object classification on the ground based on a combination of sensory modalities using binary cross-tropism as a loss function.

In today's warfare environment, software based on machine learning techniques for detecting and classifying infrastructure objects on the ground has become extremely important. In this regard, the task of improving the accuracy of classification of objects on the ground is becoming more urgent, as the use of UAVs and space systems is vital for intelligence activities. Given the nature of the input data, namely static terrain images obtained in the form of satellite images and UAV camera images, it is advisable to use convolutional neural networks to solve classification problems. In most cases, satellite images are presented in the form of multispectral and hyperspectral images, so publicly available datasets offered by the SpaceNet research community were used to train the model. An important step in preparing the training set is image orthorectification, namely adding 3D surface information to the images, which provides the model with important geometric information for semantic classes such as buildings and other structures, corrects geometric distortions, and helps the model to recognize objects in a consistent geospatial context. In the course of the experiments, the SegNet model was trained with and without the normalized Digital Surface Model (nDSM). The experimental results show that the generalized classification accuracy for six classes of objects on the test dataset increases by 23.9%. And experiments with training set limitation demonstrate that, if necessary, it is enough to use half of the available training data set to obtain only 4% lower classification accuracy and save about 10 hours of training.

New Approach Based on Pix2Pix–YOLOv7 mmWave Radar for Target Detection and Classification

Frequency modulated continuous wave (FMCW) radar is increasingly used for various detection and classification applications in different fields, such as autonomous vehicles and mining fields. Our objective is to increase the classification accuracy of objects detected using millimeter-wave radar. We have developed an approach based on millimeter-wave radar. The proposed solution combines the use of an FMCW radar, a YOLOv7 model, and the Pix2Pix architecture. The latter architecture was used to reduce noise in the heatmaps. We create a dataset of 4125 heatmaps annotated with five different object classes. To evaluate the proposed approach, 14 different models were trained using the annotated heatmap dataset. In the initial experiment, we compared the models using metrics such as mean average precision (mAP), precision, and recall. The results showed that the proposed model of YOLOv7 (YOLOv7-PM) was the most efficient in terms of mAP_0.5, which reached 90.1%, and achieved a mAP_0.5:0.95 of 49.51%. In the second experiment, we compared the models with a cleaned dataset generated using the Pix2Pix architecture. As a result, we observed improved performances, with the Pix2Pix + YOLOv7-PM model achieving the best mAP_0.5, reaching 91.82%, and a mAP_0.5:0.95 of 52.59%.

Federated Recognition Architecture Based on Voting and Feedback Mechanisms for Accuracy Object Classification in Distributed Edge Intelligence Environment

Deep learning has recently become a crucial tool to solve many complex problems and has the potential to revolutionize industries. With the widespread adoption of the Internet of Things, there are now many devices with limited computational resources that are capable of running deep learning models. It has opened up new opportunities to implement deep learning in the edge environment so that decisions can be made locally without sending data to a cloud server for processing. However, because of limited resources, the model’s performance and communication overhead are challenges when deploying learning models to edge devices. The focus of this paper is to investigate and propose a federated recognition architecture for object classification in a distributed edge intelligence environment. Specifically, we build an edge server that includes a voting module and a feedback module to improve the overall accuracy of object classification. The voting module aggregates predictions of multiple edge devices whereas the feedback module sends the voting results to edge devices to adjust the local deep learning model. We build edge devices based on the EdgeX platform which makes it easy to manage data and optimize communication overheads. Because the edge server and edge nodes only exchange prediction results, our proposed architecture ensures security with sensitive data as well as deep learning model architecture. By testing on the image dataset, we evaluate the proposed architecture’s performance and show that it outperforms individual local models in terms of accuracy. Furthermore, our experiments demonstrate that with the feedback mechanism, the deep learning model is constantly updated with new data to maintain accuracy and avoid being outdated. Besides, we prove the real-time processing speed by collecting the delay time of the proposed model. The results show that our proposed architecture has the potential to be deployed in practical applications such as smart cities, and surveillance systems.

Vehicle detection and classification using three variations of you only look once algorithm

<p><span>Vehicle detection and classification are essential for advanced driver assistance systems (ADAS) and even traffic camera surveillance. Yet, it is challenging due to complex backgrounds, varying illumination intensities, occlusions, vehicle size, and type variations. This paper aims to apply you only look once (YOLO) since it has been proven to produce high object detection and classification accuracy. There are various versions of YOLO, and their performances differ. An investigation on the detection and classification performance of YOLOv3, YOLOv4, and YOLOv5 has been conducted. The training images were from common objects in context (COCO) and open image, two publicly available datasets. The testing input images were captured on a few highways in two main cities in Malaysia, namely Shah Alam and Kuala Lumpur. These images were captured using a mobile phone camera with different backgrounds during the day and night, representing different illuminations and varying types and sizes of vehicles. The accuracy and speed of detecting and classifying cars, trucks, buses, motorcycles, and bicycles have been evaluated. The experimental results show that YOLOv5 detects vehicles more accurately but slower than its predecessors, namely YOLOv4 and YOLOv3. Future work includes experimenting with newer versions of YOLO.</span></p>

Application of YOLO v5 and v8 for Recognition of Safety Risk Factors at Construction Sites

The construction industry has high accident and fatality rates owing to time and cost pressures as well as hazardous working environments caused by heavy construction equipment and temporary structures. Thus, safety management at construction sites is essential, and extensive investments are made in management and technology to reduce accidents. This study aims to improve the accuracy of object recognition and classification that is the foundation of the automatic detection of safety risk factors at construction sites, using YOLO v5, which has been acknowledged in several studies for its high performance, and the recently released YOLO v8. Images were collected through web crawling and labeled into three classes to form the dataset. Based on this dataset, accuracy was improved by changing epochs, optimizers, and hyperparameter conditions. In each YOLO version, the highest accuracy is achieved by the extra-large model, with mAP50 test accuracies of 94.1% in v5 and 95.1% in v8. This study could be further expanded for application in various management tools at construction sites to improve the work process, quality control, and progress management in addition to safety management through the collection of more image data and automation for accuracy improvement.

Towards Indoor Suctionable Object Classification and Recycling: Developing a Lightweight AI Model for Robot Vacuum Cleaners

Robot vacuum cleaners have gained widespread popularity as household appliances. One significant challenge in enhancing their functionality is to identify and classify small indoor objects suitable for safe suctioning and recycling during cleaning operations. However, the current state of research faces several difficulties, including the lack of a comprehensive dataset, size variation, limited visual features, occlusion and clutter, varying lighting conditions, the need for real-time processing, and edge computing. In this paper, I address these challenges by investigating a lightweight AI model specifically tailored for robot vacuum cleaners. First, I assembled a diverse dataset containing 23,042 ground-view perspective images captured by robot vacuum cleaners. Then, I examined state-of-the-art AI models from the existing literature and carefully selected three high-performance models (Xception, DenseNet121, and MobileNet) as potential model candidates. Subsequently, I simplified these three selected models to reduce their computational complexity and overall size. To further compress the model size, I employed post-training weight quantization on these simplified models. In this way, our proposed lightweight AI model strikes a balance between object classification accuracy and computational complexity, enabling real-time processing on resource-constrained robot vacuum cleaner platforms. I thoroughly evaluated the performance of the proposed AI model on a diverse dataset, demonstrating its feasibility and practical applicability. The experimental results show that, with a small memory size budget of 0.7 MB, the best AI model is L-w Xception 1, with a width factor of 0.25, whose resultant object classification accuracy is 84.37%. When compared with the most accurate state-of-the-art model in the literature, this proposed model accomplished a remarkable memory size reduction of 350 times, while incurring only a slight decrease in classification accuracy, i.e., approximately 4.54%.

Monitoring method for detecting and localizing overheat, smoke and fire faults in wind turbine nacelle

This study presents a monitoring method that utilizes 3D object classification to accurately detect mechanical and electrical components of a wind turbine by combining a geometric and statistic feature extractor (GSFE) with a multiview approach. The proposed monitoring method also detect outlier after executing object detection to localize overheat faults in these components with fused Optical or Infrared/LiDAR measurements. The proposed method hasthree key characteristics. First, the proposed outlier detection allocates two extremes of normal and faulty clusters by using 2D object classification/detection model or measuring thestandard deviation of temperature with sensor fusing measurements. Specifically, the outlier detection with sensor fusing measurements extracts the position coordinates andtemperature data to localize overheat faults, effectively detecting an overheat component. Second, the GSFE utilizes a group sampling approach to extract the local geometric feature information from neighboring point clouds, aggregating normal vectors and standard deviation. This method ensures the high accuracy of object classification. Third, a multi-view approach focuses on updating local geometric and statistic features through a graph convolution network, improving the accuracy and robustness of object classification. The proposed outlier detection is verified through overheat/fire field tests. The effectiveness of the proposed 3D object classification method is also validated by using a virtual wind turbine nacelle CAD dataset and a public CAD dataset named ModelNet40. Consequently, the proposed method is practical and effective for monitoring a fire and overheat component because it can accurately detect critical components with only a few virtual datasets because gathering bigdata for training a neural network is extremely difficult.

Variational Auto-Encoder Reconstruction Networks for Classification of Hyperspectral and LiDAR Data

The classification of remote sensing scene objects has been the subject of extensive studies in recent years due to the quick advancement of earth observation and remote sensing technology. More concentration, Hyperspectral images, and LiDAR data are complementary, and their combined use for data fusion can better mine the multi-dimensional features of ground objects in remote sensing scenes, which can effectively improve the classification accuracy and reliability of ground objects in remote sensing scenes. Single-modal remote sensing data frequently cannot fully meet the needs of ground feature classification due to the increasingly complex types of ground features. In order to solve this problem, we have developed two distinct multi-source fusion classification approaches using LiDAR and hyperspectral data and deep learning techniques. They are the reconstructed multi-layer perceptron network based on the variational auto-Encoder (encoding-decoding form) and the two-stream input convolutional neural network based on the cross-channel reconstruction mechanism. These two approaches can help us find more effective and deeper feature extraction methods and feature fusion methods in this research direction and design training. We adopted the network architecture model used in this area of research and used experimental data to demonstrate the effectiveness and superiority of the suggested network model.

Visual–tactile fusion object classification method based on adaptive feature weighting

Visual–tactile fusion information plays a crucial role in robotic object classification. The fusion module in existing visual–tactile fusion models directly splices visual and tactile features at the feature layer; however, for different objects, the contributions of visual features and tactile features to classification are different. Moreover, direct concatenation may ignore features that are more beneficial for classification and will also increase computational costs and reduce model classification efficiency. To utilize object feature information more effectively and further improve the efficiency and accuracy of robotic object classification, we propose a visual–tactile fusion object classification method based on adaptive feature weighting in this article. First, a lightweight feature extraction module is used to extract the visual and tactile features of each object. Then, the two feature vectors are input into an adaptive weighted fusion module. Finally, the fused feature vector is input into the fully connected layer for classification, yielding the categories and physical attributes of the objects. In this article, extensive experiments are performed with the Penn Haptic Adjective Corpus 2 public dataset and the newly developed Visual-Haptic Adjective Corpus 52 dataset. The experimental results demonstrate that for the public dataset Penn Haptic Adjective Corpus 2, our method achieves a value of 0.9750 in terms of the area under the curve. Compared with the highest area under the curve obtained by the existing state-of-the-art methods, our method improves by 1.92%. Moreover, compared with the existing state-of-the-art methods, our method achieves the best results in training time and inference time; while for the novel Visual-Haptic Adjective Corpus 52 dataset, our method achieves values of 0.9827 and 0.9850 in terms of the area under the curve and accuracy metrics, respectively. Furthermore, the inference time reaches 1.559 s/sheet, demonstrating the effectiveness of the proposed method.

An Improving Faster-RCNN With Multi-Attention ResNet for Small Target Detection in Intelligent Autonomous Transport With 6G

Numerous object detection algorithms, such as Faster RCNN, YOLO and SSD, have been extensively applied to various fields. Both accuracy and speed of the algorithms have been significantly improved. However, as 6G technology develops, the detection effect for the small object detection task in intelligent autonomous transportation is not ideal. To strengthen the detection ability and performance of multiple scales, especially for small objects, this study proposed an object detection model based on multi-attention residual network (MA-ResNet). At first, residual network with spatial attention, channel attention, and self-attention were designed as MA-ResNet. Meanwhile, the dataset labels were smoothed. On this basis, the proposed MA-ResNet replaced the original feature extractor VGG-16 of Faster-RCNN. Different layers of MA-ResNet were extracted for feature pyramid construction. Moreover, the improved Faster-RCNN model for object detection based on MA-ResNet was formed. Furthermore, the effectiveness of the model was confirmed. The results demonstrate that MA-ResNet outperforms other feature extraction models with faster convergence speed, higher accuracy, and stronger small object classification accuracy. The improved Faster-RCNN model for object detection can effectively improve the network retrieval accuracy, performance and robustness, can exhibit satisfying adaptability for the targets of different scales in different scenarios such as vehicle identification, autonomous driving in intelligent autonomous transport system with 6G. Moreover, the study provides a certain reference for improving the effect of multiple scales, especially small object detection in intelligent transportation with 6G.

The Design of a New 3D Print-in-place Soft Four-Legged Robots with Artificial Intelligence

Soft and flexible robots are designed to change their flexibility over a wide range to perform tasks adequately in real-world applications. Current soft robots require cast moulding, high assembly effort and large actuators. Soft origami structures exhibit high levels of compliance. In this paper, we designed a new 3D print-in-place soft four-legged robot (3DSOLR). Our soft legged robot is an endurance application adapted from the soft origami zigzag gripper. This novel and innovative design are inspired by the rigid joint Theo Jansen legged robot with highly adaptive 3D print-in-place soft origami legs capable of fluid motion and even surviving drop tests. The robot mechanism consists of four soft origami flexible legs driven by two DC motors. The 3DSOLR is lightweight and semi-autonomous using two Hall effect sensors and a wireless Bluetooth module. Being 3D print-in-place using Thermoplastic polyurethane also increases its durability while having flexibility, simplicity and safety. The robot also has a gripper inspired by the mandible of male European stag beetle (Lucanus cervus). These features make this robot suitable to be used in social robotics and rescue robotics applications. The transmitter program is implemented in Bluetooth serial communication using MIT App Inventor 2 smartphone apps and a microcontroller Arduino ATMEL is used as the main controller and code in Arduino IDE. It has artificial intelligence (AI) capability with ESP32 CAM onboard which has an object classification accuracy of 95.5% using custom Edge Impulse neural network MobileNetV1 96 x 96. This AI capability enhanced the robot’s capability in object classification for grasping.

Research of the Characteristics of a Convolutional Neural Network on the ESP32-CAM Microcontroller

The paper is devoted to solving the problem of using neural networks for real-time image recognition on low-power portable devices running on microcontrollers. The ESP-32© CAM microcontroller was used as the target device, on which an artificial neural network was deployed, written using the Python® programming language and the Tensorflow© library for building neural networks. The performance of the microcontroller and personal computer for object detection using a neural network and their classification were compared in the paper. The image recognition time and percentage of correctly classified objects were compared. The paper shows that the number of training epochs affects the accuracy of object classification in the image. The obtained results show that increasing the number of training epochs increases the accuracy of object recognition using the studied neural network, but a significant increase in the number of epochs does not lead to a significant improvement in recognition accuracy. The difference in the obtained results for the microcontroller and personal computer image recognition accuracy ranges from 5%.