Image Classification Technology Research Articles

SA-ConvNeXt: A Hybrid Approach for Flower Image Classification Using Selective Attention Mechanism

In response to the current lack of annotations for flower images and insufficient focus on key image features in traditional fine-grained flower image classification based on deep learning, this study proposes the SA-ConvNeXt flower image classification model. Initially, in the image preprocessing stage, a padding algorithm was used to prevent image deformation and loss of detail caused by scaling. Subsequently, the model was integrated using multi-level feature extraction within the Efficient Channel Attention (ECA) mechanism, forming an M-ECA structure to capture channel features at different levels; a pixel attention mechanism was also introduced to filter out irrelevant or noisy information in the images. Following this, a parameter-free attention module (SimAM) was introduced after deep convolution in the ConvNeXt Block to reweight the input features. SANet, which combines M-ECA and pixel attention mechanisms, was employed at the end of the module to further enhance the model’s dynamic extraction capability of channel and pixel features. Considering the model’s generalization capability, transfer learning was utilized to migrate the pretrained weights of ConvNeXt on the ImageNet dataset to the SA-ConvNeXt model. During training, the Focal Loss function and the Adam optimizer were used to address sample imbalance and reduce gradient fluctuations, thereby enhancing training stability. Finally, the Grad-CAM++ technique was used to generate heatmaps of classification predictions, facilitating the visualization of effective features and deepening the understanding of the model’s focus areas. Comparative experiments were conducted on the Oxford Flowers102 flower image dataset. Compared to existing flower image classification technologies, SA-ConvNeXt performed excellently, achieving a high accuracy of 96.7% and a recall rate of 98.2%, with improvements of 4.0% and 3.7%, respectively, compared to the original ConvNeXt. The results demonstrate that SA-ConvNeXt can effectively capture more accurate key features of flower images, providing an effective technical means for flower recognition and classification.

Advancing Fruit Image Classification with State-of-the-Art Deep Learning Techniques

Fruit image classification technology using deep learning is making significant contributions in the agriculture and food retail sectors, promising to increase efficiency and productivity. However, there is an identified knowledge gap in dealing with the considerable variation in fruit appearance caused by factors such as type, size, color, and lighting conditions, as well as the precise identification of damage or disease. This research focuses on applying the developed Convolutional Neural Network architecture to fill this gap by using it in an extensive and diverse dataset, covering 67,692 image files categorized into 131 fruit classes. The training process showed substantial accuracy improvement, with training accuracy reaching 98.39% and validation accuracy at 90%, while training loss decreased to 0.0430 and validation loss to 0.2991. In the advanced stage of training, the training accuracy peaked at 99.43% in the 59th epoch with a shallow loss of 0.0251. However, the validation loss showed variation, indicating room for improvement in model generalization. The findings provide insight into the potential and challenges of applying Convolutional Neural Network models and fruit image classification with improved fruit sorting accuracy. Contribution to the literature in the field of information technology and agriculture by showing deep learning models can be improved to address the issue of fruit image variability.

A study on fruit classification using convolutional neural network for image recognition technology

Artificial intelligence is being incorporated into more aspects of our daily lives. Agricultural production in China is at a critical juncture due to the rising fruit production and dwindling labor force, necessitating the adoption of mechanization and intelligent systems. Image processing is particularly suitable for this field. Image classification technologies in fruit categorization are examined in this research. The results of this study can help improve the development of intelligent, lightweight machinery for agricultural output. This, in turn, will enhance the efficiency of fruit cultivation, harvesting, and trading and alleviate labor constraints. Two popular convolutional neural network models, namely ResNet50 and MobileNetV2, were employed in this study. The study utilized two optimizers: SGD and Adam. The evaluation results revealed that the ResNet50 model, employing SGD optimization, achieved the highest accuracy of 95.57%. Despite its lower accuracy of 92.19%, the MobileNetV2 model demonstrates higher efficiency than ResNet50 due to its lower hardware requirements, rendering it suitable for operation on compact devices.

Fire image detection and classification analysis based on VGG16 image processing model

Fire image classification technology refers to the classification and recognition of fire images through computer vision technology in order to take timely countermeasures. With the development of computer vision technology, fire image classification technology has been widely studied and applied. Deep learning techniques have achieved great success in the field of image classification, with researchers classifying and identifying fire images by using deep learning models such as convolutional neural networks. This paper introduces the research background and application of fire image classification technology, and the experimental results of detection and classification analysis of fire image based on vgg16 image processing model. The model can classify and identify fire images well and achieve good prediction effect. The accuracy of training set and test set is stable at 99%, the accuracy and accuracy of the model reach 98%, and the recall rate and F1 score reach 99%. The application of fire image classification technology is of great significance. By using computer vision technology to classify and identify fire images, it can improve the accuracy and efficiency of fire monitoring and early warning system, timely detect and control fire, and reduce the loss caused by fire. At the same time, the continuous development of deep learning technology also provides a broader space for the research and application of fire image classification technology.

Image defect classification of surface mount technology welding based on the improved ResNet model

In mass production, welding flaw detection in existing surface mount technology (SMT) has certain constraints, including its high costs, heavy workloads, and time-consuming processes. However, image classification technology using computer vision demonstrates high detection speeds and considerably reduced detection costs in flaw detection. Nevertheless, the increased integration of chip components on printed circuit boards (PCBs) and reduced component sizes pose challenges for flaw detection technology. Therefore, in this paper, an SMT welding image flaw classification model—that is, the ResNet-34-ECA model—based on an improved ResNet model, is proposed. Initially, the dataset is amplified using data amplification methods, such as stochastic rotation, increased data diversity, and enhanced model robustness. The ResNet34 model is then optimized using the light quantization efficient channel attention (ECA) module, resulting in higher classification accuracy. The experimental data in this study were collected using automated optical inspection (AOI) equipment, following the manual creation and amplification of the dataset. The experimental results showed that the baseline model accuracy increased by 0.22 in the augmented dataset, reaching 97.2%. Moreover, the ResNet-34-ECA model proposed in this paper could realize the classification of SMT welding image defects successfully; the overall classification accuracy of the improved ResNet image classification model was 0.01 higher than that of the baseline model, reaching 98.2%. Consequently, the proposed model proves to be better than other models in defect classification on this dataset, providing an accurate classification of SMT welding image defects.

Research on the model of automatic recognition and natural language question-answer system for traditional Chinese medicine tongue images based on LLMs

Large Language Models (LLMs) have recently demonstrated their potential in clinical applications by providing valuable medical knowledge and recommendations. Traditional Chinese tongue diagnosis, one of the "Four Diagnoses," is an essential method for traditional Chinese medicine diagnosis. This paper builds upon tongue image classification technology and utilizes natural language processing and image recognition techniques to enhance the discrimination and analysis of traditional Chinese tongue images through learning and inference. We propose a method to integrate LLMs into the tongue image automatic recognition model and use them in interactive question-answering by summarizing and reorganizing information in natural language text format.

Predicting the ash content in coal flotation concentrate based on convolutional neural network

ABSTRACT Convolutional neural networks (CNNs) are currently one of the most popular image classification technologies. Their excellent image classification ability enables the prediction of the ash content in clean coal using coal flotation froth images. Herein, a scheme is proposed to predict the ash content in coal flotation concentrate using coal flotation froth images and CNN architecture. According to the ash content of the concentrate, the coal flotation froth image data set is divided into seven interval categories, and data augmentation is used to expand the data set. Then, several recently trained CNNs (ResNet, EfficientNet, EfficientNetV2, ConvNeXt) are used to classify the coal flotation froth images with different concentrate ash content interval (integer ±1%). The classification performance of each network model, relationship between model performance and hyperparameters, and the abstract pixel features of the best model are visualized. The EfficientNetV2-L network achieves the highest classification accuracy (99%) after fine-tuning. The results indicate that the CNNs are effective in predicting the ash content of clean coal by classifying coal flotation froth images. The optimized CNN model is applied to froth images in the industry, resulting in high classification performance, which demonstrates that the CNN has high potential for industrial applications.

CLASSIFICATION OF DISTRACTED DRIVER USING SUPPORT VECTOR MACHINE BASED ON PRINCIPAL COMPONENT ANALYSIS FEATURE REDUCTION AND CONVOLUTIONAL NEURAL NETWORK

The use of ground transportation in Indonesia, especially in major cities like Surabaya, has experienced rapid growth. However, this increased usage has also led to a rise in traffic accidents. One of the main contributing factors is driver distraction. Therefore, this study aims to develop a method for detecting distracted drivers using image classification technology with Convolutional Neural Networks (CNN) and Support Vector Machine (SVM) models. In this research, data was obtained from the "State Farm Distracted Driver Detection" dataset, which contains images of drivers who are either distracted or not focused while driving. The initial process involves data preprocessing, such as resizing images to 50 x 50 pixels and dividing the dataset into training and testing data. Next, feature extraction is performed using a CNN model with three convolutional layers, three Maxpooling layers, and one flattened layer. After feature extraction, the Principal Component Analysis (PCA) method is used to reduce the dimensionality of the data. Furthermore, an SVM model was trained using data reduced by PCA with a 60:40 data split. This research conducted a comparison between the use of PCA and not using PCA. Based on the test results, the use of PCA not only improved the classification accuracy to 96.28% compared to 92.46% without PCA but also accelerated the training time to 10.64 seconds from 19.67 seconds without PCA.

Studies Advanced in Image Classification based on Deep Learning

Accurate and efficient image classification is one of the important research topics in image analysis, and it has also been a research hotspot in the computer vision community.Deep neural networks are increasingly being used for picture categorization and processing in recent years as a result of advancements in machine learning technology.In this article, we introduce the research progress of image recognition technology based on depth learning, including the design ideas, principles, structures, advantages and disadvantages of several depth neural networks. In addition, we quantitatively compared the recognition results of different methods on classic image classification datasets. Lastly, we review some current issues with picture classification and talk about the direction of deep neural network development in the use of image classification technology.

A Design of Bare Printed Circuit Board Defect Detection System Based on YOLOv8

As electronic products develop towards miniaturization and digitization, printed circuit boards (PCBs) also develop towards high density and high precision. In the manufacturing process of PCBs, some PCBs with defects will be produced, and these defects often lead to circuit failure, so defect detection technology is an indispensable part of PCB manufacturing technology. Aiming at the problems of low efficiency and accuracy of traditional image recognition and classification technology, A PCB defect detection algorithm is proposed based on YOLOv8 in this paper. For these five PCB defects, the neural network in deep learning was used to identify and classify PCB defects. The prediction accuracy of YOLOv8 model after training is close to 97%, and it is compared with the accuracy of other algorithms to prove the effectiveness and feasibility of the model. In addition, the object detection image user interface is also established in this paper, which can realize image detection more conveniently.

Automatic Classification Method for Oracle Images based on Deep Learning

There are some problems in oracle image classification, such as high image-feature extraction error and low classification accuracy, which leads to low efficiency. Therefore, we designed an automatic classification method for oracle images based on deep learning. The clustering center of an oracle image is determined using k-means algorithm. The obtained oracle image block is used as parallel input to extract a dictionary. The oracle image features are output through a feature mapping function for feature extraction. The CS algorithm is used to reduce the dimension of the oracle image, and the oracle feature compressed sensing signal is sparse under a certain basis for preprocessing of image features. The center of six feature points in the features is used to describe the center of the feature points of the whole oracle image, to determine the position change state of the feature points, and to characterize the key features of the image to achieve the target alignment of oracle image feature points. Based on this, the image feature data are learned using a transfer learning model. The automatic classifier was constructed through a convolution neural network. Experimental results show that the proposed method can effectively reduce the error of image feature extraction and improve the classification accuracy. At the application level, the research results of this paper have brought great improvement to the performance of oracle image classification technology. At the theoretical level, this research introduces k-means and CS algorithms to oracle image classification technology based on deep learning and makes an effective exploration in the direction of this technological progress.

Max–min distance embedding for unsupervised hyperspectral image classification in the satellite Internet of Things system

In modern society, hyperspectral images have widely been used in many applications. In the earth observation system, hyperspectral images are acquired by hyperspectral sensors and transmitted to the cloud computing center for analysis, which are the part of satellite Internet of Things(IoT) to build a large scale land-cover surveillance system. Thus, during recent years, researchers throughout academia and industry have been advancing the classification of hyperspectral images. Luckily for swamped remote sensing researchers, new technology is allowing brands to extract useful knowledge using artificial intelligence to enhance supervised hyperspectral image classification technology. However, conventional methods that label and process hyperspectral images manually are labor and time-consuming. Hence, in this paper, we proposed an unsupervised hyperspectral image classification framework. The paper differs from the existing methods by three aspects: (1) We adopt multi-scale spatial features to enhance the discriminability of data. (2) We propose deep autoencoder with max–min distance embedding to extract features and reduce dimensions of data, which is more suitable for large-scale data clustering. (3) We apply k-means clustering method to get original classification, then optimize the results by the guided filter. Experimental results on two widely used hyperspectral data, Salinas and Pavia University, show the competitive performance obtained by the proposed framework compared with other approaches.

Image Classification Under Class-Imbalanced Situation

Image classification technology processes and analyzes image data to extract valuable feature information to distinguish different types of images, thereby completing the process of machine cognition and understanding of image data. As the cornerstone of image application field, image classification technology involves a wide range of application fields. The class imbalance distribution is ubiquitous in the application of image classification and is one of the main problems in image classification research. This study summarizes the literature on class-imbalanced image classification methods in recent years, and analyzes the classification methods from both the data level and the algorithm level. In data-level methods, oversampling, under sampling and mixed sampling methods are introduced, and the performance of these literature algorithms is summarized and analyzed. The algorithm-level classification method is introduced and analyzed from the aspects of classifier optimization and ensemble learning. All image classification methods are analyzed in detail in terms of advantages, disadvantages and datasets.

Urban Tree Canopy and Environmental Justice: Examining the Distributional Equity of Urban Tree Canopy in Guangzhou, China.

Rapid urbanization has led to a series of environmental problems that are harmful to the physical and mental health of residents. Not only can increasing urban tree canopy (UTC) improve sustainable city development, but it can also effectively promote the quality of life for residents; however, the uneven spatial distribution of UTC can also bring about social justice problems. Currently, there are few studies related to the justice of UTC distribution in China. Based on this, the paper adopts object-oriented image classification technology to interpret and extract UTC data from satellite images, and it uses house price to explore the justice of the spatial distribution of UTC in the main urban area of Guangzhou from the perspective of environmental justice by ANOVA, Pearson correlation coefficient, and bivariate local spatial autocorrelation. The results show that: (1) There is a significant positive correlation between UTC and house price in the main urban area of Guangzhou, and there are regional differences in the distribution of UTC, with the UTC in the "very high" house price group being significantly higher than that in other groups. (2) The spatial clustering of UTC and house price in the main urban area of Guangzhou is found to be "low-low" and "high-high"; thus, it can be concluded that the spatial distribution of UTC in the main urban area of Guangzhou is uneven. This is an environmental injustice, as the areas with low UTC are spatially clustered in old residential areas, while the areas with high UTC are spatially clustered in commercial housing estates with high house prices. The study shows that urban tree planting should not only focus on quantitative improvement but also on equitable spatial layouts to promote social equity and justice thereby improving the urban ecological environment and promoting healthy urban development.

CLASSIFICATION OF BATIK MOTIF USING TRANSFER LEARNING ON CONVOLUTIONAL NEURAL NETWORK (CNN)

The number of batik motifs in Indonesia is not comparable to the knowledge possessed by the Indonesian people about batik motifs. The diversity of batik motifs can be a problem because classifying them can only be done by those who are familiar with batik in depth, both the pattern and the philosophy behind the motif, most of which are elderly people. To classify batik accurately and quickly is to use image classification technology. In this study, data were obtained from the previous researchers' GitHub repository, google images, and camera shots with a total dataset of 3,534 images. The data only focused on five batik motifs, namely Ceplok, Kawung, Parang, Megamendung, and Sidomukti. Before the batik motif is processed, preprocessing is carried out to obtain various quality data. Then the dataset was trained using the CNN model then the results were retrained using the VGG-16 and Xception Transfer Learning models. The researcher made several model scenarios, namely the CNN model without Transfer Learning and the model with Transfer Learning which took into account the effect of the learning rate values ​​of 0.0004 and 0.0001. Therefore, the results of the CNN model without Transfer Learning (M0) obtained training accuracy results of 89.64%. While the results of the model with the best Transfer Learning is the M2 model (CNN + VGG-16, learning rate = 0.0001) with an accuracy of 91.23%, a loss of 24.48%, and the test results obtained an accuracy of 89%. Based on the results of the classification method, it can be concluded that the CNN model with Transfer Learning performs classification better in terms of accuracy and computation time than the CNN model.

Shape classification technology of pollinated tomato flowers for robotic implementation

Three pollination methods are commonly used in the greenhouse cultivation of tomato. These are pollination using insects, artificial pollination (by manually vibrating flowers), and plant growth regulators. Insect pollination is the preferred natural technique. We propose a new pollination method, using flower classification technology with Artificial Intelligence (AI) administered by drones or robots. To pollinate tomato flowers, drones or robots must recognize and classify flowers that are ready to be pollinated. Therefore, we created an AI image classification system using a machine learning convolutional neural network (CNN). A challenge is to successfully classify flowers while the drone or robot is constantly moving. For example, when the plant is shaking due to wind or vibration caused by the drones or robots. The AI classifier was based on an image analysis algorithm for pollination flower shape. The experiment was performed in a tomato greenhouse and aimed for an accuracy rate of at least 70% for sufficient pollination. The most suitable flower shape was confirmed by the fruiting rate. Tomato fruit with the best shape were formed by this method. Although we targeted tomatoes, the AI image classification technology is adaptable for cultivating other species for a smart agricultural future.

Multifrequency PolSAR Image Fusion Classification Based on Semantic Interactive Information and Topological Structure

Compared with the rapid development of single-frequency polarimetric SAR (PolSAR) image classification technology, there is less research on the land cover classification of multi-frequency PolSAR (MF-PolSAR) images. And the deep learning methods among them are mainly based on convolutional neural networks (CNNs), only local spatiality is considered but the nonlocal relationship is ignored. Therefore, this paper proposes the MF semantics and topology fusion (MF-STF) model based on semantic interaction and nonlocal topological structure to improve MF-PolSAR classification performance. During MF-STF optimization, the semantic information-based classification (SIC) and topological property-based classification (TPC) work collaboratively, not only fully leveraging the complementarity of bands, but also combining local and nonlocal spatial information to improve the discrimination of different categories. For SIC, the designed cross-band interactive feature extraction (CIFE) module is embedded to explicitly model the deep semantic correlation among bands, thereby leveraging the complementarity of bands to make ground objects more separable. In TPC, the graph sample and aggregate network (GraphSAGE) is employed to dynamically capture the representation of nonlocal topological relations between land cover categories. In this way, the robustness of classification can be further improved by combining nonlocal spatial information. Finally, a MF weighted fusion (MFWF) strategy is proposed to merge inference from different bands, so as to make the MF joint classification decisions of SIC and TPC. Notably, its weights are adjusted based on the total model loss. The effectiveness of the proposed modules is proved by ablation experiments on three measured MF-PolSAR datasets. In addition, the comparative experiments show that MF-STF can achieve more competitive classification performance than some state-of-the-art methods.

Monkeypox Classification based on Skin Images using CNN: EfficientNet-B0

Monkeypox is a zoonotic infectious disease caused by a virus of the orthopoxvirus genus. It can infect humans, vertebrates, and arthropods. Transmission to humans occurs through direct contact with infected animal body fluids or consumption of undercooked meat. Monkeypox cases have been reported globally, with thousands of confirmed cases and several deaths. Early symptoms include fever, rash, swollen lymph nodes, back pain, and headache. Diagnosis can be made through physical examination and laboratory tests. Imagebased artificial intelligence technology, specifically the EfficientNet-B0 architecture, has been proposed as a solution for the classification of monkeypox based on skin lesion images. The research aims to compare the performance of EfficientNetB0 with other CNN architectures and contribute to the development of medical image classification technology. Among the models evaluated, the EfficientNet-B0 model emerged as the standout performer, achieving an accuracy of 85.12%, surpassing the accuracy of other models such as MobileNet (63.63%) and InceptionV3 (71.4%). EfficientNet-B0 also demonstrated strong sensitivity (78.46%) and impressive specificity (91.78%), outperforming other models in these metrics. Additionally, despite not surpassing the accuracy of ResNet-50 (87.59%), EfficientNet-B0 achieved its accuracy with approximately four times fewer parameters, highlighting its efficiency in parameter usage and computational resources. These results can help improve models and aid in clinical decision-making.

Research on cassava disease classification using the multi-scale fusion model based on EfficientNet and attention mechanism.

Cassava disease is one of the leading causes to the serious decline of cassava yield. Because it is difficult to identify the characteristics of cassava disease, if not professional cassava growers, it will be prone to misjudgment. In order to strengthen the judgment of cassava diseases, the identification characteristics of cassava diseases such as different color of cassava leaf disease spots, abnormal leaf shape and disease spot area were studied. In this paper, deep convolutional neural network was used to classify cassava leaf diseases, and image classification technology was used to recognize and classify cassava leaf diseases. A lightweight module Multi-scale fusion model (MSFM) based on attention mechanism was proposed to extract disease features of cassava leaves to enhance the classification of disease features. The resulting feature map contained key disease identification information. The study used 22,000 cassava disease leaf images as a data set, including four different cassava leaf disease categories and healthy cassava leaves. The experimental results show that the cassava leaf disease classification model based on multi-scale fusion Convolutional Neural Network (CNN) improves EfficientNet compared with the original model, with the average recognition rate increased by nearly 4% and the average recognition rate up to 88.1%. It provides theoretical support and practical tools for the recognition and early diagnosis of plant disease leaves.

Segmentation Effect on Lungs X-Ray Image Classification Using Convolution Neural Network

The effect of segmentation on lung X-ray image classification has been analyzed in this study. The 150 lung x-ray images in this study were separated into 78 as training data, 30 as validation data, and 42 as testing in three categories: normal lungs, effusion lungs, and cancer lungs. In pre-processing, the images were modified by adaptive histogram equalization to improve image quality and increase image contrast. The segmentation aims to mark the image by contouring the lung area obtained from the thresholding and some morphological manipulation processes such as filling holes, area openings, and labelling. Image classification uses Convolutional Neural Network (CNN) with five convolution layers, an Adam optimizer, and 30 epochs. The segmentation effect is analyzed by comparing the classification performance of the segmented and unsegmented images. In the study, the unsegmented X-ray image dataset classification reached an overall accuracy of 59.52% in the network testing process. The segmented X-ray image dataset obtained greater accuracy, 73.81%. It indicated that the segmentation process could improve network performance because the input pattern of the segmented image is easier to classify. Furthermore, the segmentation technique in the study can be one of the alternatives to developing image classification technologies, especially for medical image diagnosis. Segmentation Effect on Lungs X-Ray Image Classification Using Convolution Neural Network.