Image Classification Method Research Articles

CATNet: Cascaded attention transformer network for marine species image classification

Complex physicochemical environmental effects result in the underwater species images’ highly intricate and diverse backgrounds, which poses significant challenges for identifying marine species. Deep learning techniques are used extensively for image classification thanks to their exceptional feature extraction capabilities. To address the above issues, we present a cascaded attention transformer network for marine species image classification named CATNet. More specifically, we first utilize the efficient channel attention module with depthwise convolution, which can efficiently extract the essential features of the image and reduce the computational parameters of the traditional attention module. Subsequently, we design a stacked transformer module to focus on the different features among a large amount of feature information, thereby reducing redundancy and improving?the generalization performance of CATNet. To comprehensively exploit the features extracted from different modules, we integrate the efficient channel attention and stacked transformer modules to construct a feature-cascaded module. Meanwhile, we construct a large-scale Marine Species Image Dataset (MSID) including 8605 underwater images with different species. Concretely, it consists of 712 coral reef images, 1270 fish images, 1082 people images, 2414 sea urchin images, 1219 squid images, 837 sea cucumber images, and 1071 turtle images. Comprehensive experimentation on our built dataset underscores the proposed CATNet’s superiority over the existing state-of-the-art methods in marine species image classification. The data is available. https://www.researchgate.net/publication/381961074_2024MSID.

TST_MFL: Two-stage training based metric fusion learning for few-shot image classification

Addressing the limitations of most few-shot learning (FSL) methods, particularly their insufficient single-feature discriminability and generalization during pre-training encoding, this paper introduces a novel approach: Two-Stage Training based Metric Fusion Learning (TST_MFL) for few-shot image classification. TST_MFL innovatively integrates a two-stage fusion metric learning strategy, enhancing both pre-training and meta-training phases to overcome the prevalent focus on single-level feature extraction in current few-shot image classification methods. In pre-training stage, we employ prediction distillation loss and local difference loss to facilitate mutual learning between global and local features, thus obtaining more discriminative features. Moreover, meta-training further enhances feature generalization and integrates global–local metrics to improve few-shot classification accuracy. Our approach demonstrates competitive performance across three fine-grained datasets and one coarse-grained dataset. Notably, experimental results in 1-shot and 5-shot settings of fine-grained datasets underscore the superiority of TST_MFL. The code for this work is available at: https://github.com/ZitZhengWang/TST_MFL.

Classification Method of Composite Insulator Surface Image Based on GAN and CNN

Classification Method of Composite Insulator Surface Image Based on GAN and CNN

DEEP NEURAL MULTI WAVELET SEGMENTATION AND RESNET-50 IMAGE CLASSIFICATION (DNMWS-RESNET-50IC) BASED EARLY GLAUCOMA DETECTION

Glaucoma, a leading cause of irreversible blindness worldwide, results from progressive damage to the optic nerve, often linked to elevated Intra Ocular Pressure (IOP). Early detection is critical for preventing vision loss, yet traditional diagnostic methods can be limited in accessibility and effectiveness, particularly in the early stages of the disease. Addressing the need for early detection, we propose the Deep Neural Multi-Wavelet Segmentation and ResNet-50 Image Classification (DNMWS-ResNet-50IC) method, specifically designed to detect glaucoma at an early stage using retinal fundus images. In this work, we apply Anisotropic Gaussian Filtering for the preprocessing of retinal fundus images, effectively reducing image noise while preserving the original quality of the image pixels by creating an adaptive window size and scale space. We then utilize multi- wavelet-based image segmentation, leveraging wavelet transforms to analyze and decompose the image into its various frequency components. This technique is particularly advantageous for managing images with complex structures and textures. Subsequently, the segmented features are classified using the ResNet-50 model, which categorizes the images as normal, abnormal, or indicative of early- stage glaucoma. The effectiveness of the proposed method is assessed by measuring three key performance indicators sensitivity, specificity, and accuracy on digital retinal images from the HRF image database. Additionally, the model's performance is further evaluated on a separate test set, considering metrics such as accuracy, precision, recall, and prediction time.

Border-Enhanced Triple Attention Mechanism for High-Resolution Remote Sensing Images and Application to Land Cover Classification

With the continuous development and popularization of remote sensing technology, remote sensing images have been widely used in the field of land cover classification. Since remote sensing images have complex spatial structure and texture features, it is becoming a challenging problem to accurately categorize them. Land cover classification has practical application value in various fields, such as environmental monitoring and protection, urban and rural planning and management, and climate change research. In recent years, remote sensing image classification methods based on deep learning have been rapidly developed, in which semantic segmentation technology has become one of the mainstream methods for land cover classification using remote sensing image. Traditional semantic segmentation algorithms tend to ignore the edge information, resulting in poor classification of the edge part in land cover classification, and there are numerous attention mechanisms to make improvements for these problems. In this paper, a triple attention mechanism, BETAM (Border-Enhanced Triple Attention Mechanism), for edge feature enhancement of high-resolution remote sensing images is proposed. Furthermore, a new model on the basis of the semantic segmentation network model DeeplabV3+ is also introduced, which is called DeepBETAM. The triple attention mechanism BETAM is able to capture feature dependencies in three dimensions: position, space, and channel, respectively. Through feature importance weighting, modeling of spatial relationships, and adaptive learning capabilities, the model BETAM pays more attention to edge features, thus improving the accuracy of edge detection. A remote sensing image dataset SMCD (Subject Meticulous Categorization Dataset) is constructed to verify the robustness of the attention mechanism BETAM and the model DeepBETAM. Extensive experiments were conducted on the two self-built datasets FRSID and SMCD. Experimental results showed that the mean Intersection over Union (mIoU), mean Pixel Accuracy (mPA), and mean Recall (mRecall) of DeepBETAM are 63.64%, 71.27%, and 71.31%, respectively. These metrics are superior to DeeplabV3+, DeeplabV3+(SENet), DeeplabV3+(CBAM), DeeplabV3+(SAM), DeeplabV3+(ECANet), and DeeplabV3+(CAM), which are network models that incorporate different attention mechanisms. The reason is that BETAM has better edge segmentation results and segmentation accuracy. Meanwhile, on the basis of the self-built dataset, the four main classifications of buildings, cultivated land, water bodies and vegetation were subdivided and detected, and good experimental results were obtained, which verified the robustness of the attention mechanism BETAM and the model DeepBETAM. The method has broad application prospects and can provide favorable support for research and application in the field of surface classification.

A medical image classification method based on self-regularized adversarial learning.

Deep learning (DL) techniques have been extensively applied in medical image classification. The unique characteristics of medical imaging data present challenges, including small labeled datasets, severely imbalanced class distribution, and significant variations in imaging quality. Recently, generative adversarial network (GAN)-based classification methods have gained attention for their ability to enhance classification accuracy by incorporating realistic GAN-generated images as data augmentation. However, the performance of these GAN-based methods often relies on high-quality generated images, while large amounts of training data are required to train GAN models to achieve optimalperformance. In this study, we propose an adversarial learning-based classification framework to achieve better classification performance. Innovatively, GAN models are employed as supplementary regularization terms to support classification, aiming to address the challenges describedabove. The proposed classification framework, GAN-DL, consists of a feature extraction network (F-Net), a classifier, and two adversarial networks, specifically a reconstruction network (R-Net) and a discriminator network (D-Net). The F-Net extracts features from input images, and the classifier uses these features for classification tasks. R-Net and D-Net have been designed following the GAN architecture. R-Net employs the extracted feature to reconstruct the original images, while D-Net is tasked with the discrimination between the reconstructed image and the original images. An iterative adversarial learning strategy is designed to guide model training by incorporating multiple network-specific loss functions. These loss functions, serving as supplementary regularization, are automatically derived during the reconstruction process and require no additional dataannotation. To verify the model's effectiveness, we performed experiments on two datasets, including a COVID-19 dataset with 13958 chest x-ray images and an oropharyngeal squamous cell carcinoma (OPSCC) dataset with 3255 positron emission tomography images. Thirteen classic DL-based classification methods were implemented on the same datasets for comparison. Performance metrics included precision, sensitivity, specificity, and -score. In addition, we conducted ablation studies to assess the effects of various factors on model performance, including the network depth of F-Net, training image size, training dataset size, and loss function design. Our method achieved superior performance than all comparative methods. On the COVID-19 dataset, our method achieved , , , and in terms of precision, sensitivity, specificity, and -score, respectively. It achieved across all these metrics on the OPSCC dataset. The study to investigate the effects of two adversarial networks highlights the crucial role of D-Net in improving model performance. Ablation studies further provide an in-depth understanding of ourmethodology. Our adversarial-based classification framework leverages GAN-based adversarial networks and an iterative adversarial learning strategy to harness supplementary regularization during training. This design significantly enhances classification accuracy and mitigates overfitting issues in medical image datasets. Moreover, its modular design not only demonstrates flexibility but also indicates its potential applicability to various clinical contexts and medical imaging applications.

MM_Fast_RCNN_ResNet: Construction of Multimodal Faster RCNN Inception and ResNet V2 for Pedestrian Tracking and detection

Pedestrian identification and tracking is a crucial duty in smart building monitoring. The development of sensors has led to architects' focus on smart building design. The image distortions caused by numerous external environmental factors present a significant problem for pedestrian recognition in smart buildings. It is difficult for machine learning algorithms and other conventional filter-based image classification methods, such as histograms of oriented gradient filters, to function efficiently when dealing with many input photos of pedestrians. Deep learning algorithms are now performing substantially better when processing an enormous amount of image data. This article evaluates a novel multimodal classifier-based pedestrian identification method. The proposed method is Multimodal Faster RCNN Inception and ResNet V2 (MM Fast RCNN ResNet). The collected attributes address a tracking problem and establish the foundation for several object recognition tasks (novelty). Our method's neural network is regularized, and the feature representation is automatically adjusted to the detection assignment, resulting in high accuracy (superior to the proposed method). The proposed method is assessed using the PenFudan dataset and contemporary techniques regarding several factors. It is discovered that the recommended MM Fast RCNN ResNet obtains precision, recall, FPPI, FPPW, and average precision of 0.9057, 0.8629, 0.0898, and 0.0943.

Automatic Quality Assessment of Pork Belly via Deep Learning and Ultrasound Imaging.

Pork belly, prized for its unique flavor and texture, is often overlooked in breeding programs that prioritize lean meat production. The quality of pork belly is determined by the number and distribution of muscle and fat layers. This study aimed to assess the number of pork belly layers using deep learning techniques. Initially, semantic segmentation was considered, but the intersection over union (IoU) scores for the segmented parts were below 70%, which is insufficient for practical application. Consequently, the focus shifted to image classification methods. Based on the number of fat and muscle layers, a dataset was categorized into three groups: three layers (n = 1811), five layers (n = 1294), and seven layers (n = 879). Drawing upon established model architectures, the initial model was refined for the task of learning and predicting layer traits from B-ultrasound images of pork belly. After a thorough evaluation of various performance metrics, the ResNet18 model emerged as the most effective, achieving a remarkable training set accuracy of 99.99% and a validation set accuracy of 96.22%, with corresponding loss values of 0.1478 and 0.1976. The robustness of the model was confirmed through three interpretable analysis methods, including grad-CAM, ensuring its reliability. Furthermore, the model was successfully deployed in a local setting to process B-ultrasound video frames in real time, consistently identifying the pork belly layer count with a confidence level exceeding 70%. By employing a scoring system with 100 points as the threshold, the number of pork belly layers in vivo was categorized into superior and inferior grades. This innovative system offers immediate decision-making support for breeding determinations and presents a highly efficient and precise method for assessment of pork belly layers.

Industry Image Classification Based on Stochastic Configuration Networks and Multi-Scale Feature Analysis.

For industry image data, this paper proposes an image classification method based on stochastic configuration networks and multi-scale feature extraction. The multi-scale features are extracted from images of different scales using deep 2DSCN, and the hidden features of multiple layers are also connected together to obtain more informational features. The integrated features are fed into SCNs to learn a classifier which improves the recognition rate for different categories. In the experiments, a handwritten digit database and an industry hot-rolled steel strip database are used, and the comparison results demonstrate the proposed method can effectively improve the classification accuracy.

LCFE-X: Pluggable Local Channel Feature Extraction model enhancement method for hyperspectral image classification

Spectral information of hyperspectral images offers high-dimensional features for downstream tasks. Local spectral signatures in different wavelength ranges are critical for material identification and pixel classification. In line with conventional vision methods, recent studies have focused on bolstering the spatial modelling capabilities of transformers. However, the focus remains predominantly global when the channel dimension is considered. Current research on enhancing local channel feature extraction is primarily model specific and lacks generalisable methods that can be applied to various models. Therefore, this study introduces a Local Channel Feature Extraction (LCFE) module. It extracts spectral-spatial features and key features within each band group and then integrates the features to obtain local channel features. Subsequently, a pluggable LCFE-X model enhancement method is proposed. This involves performing LCFE at the front end of the model and fusing the extracted features with the original features. Three enhanced hyperspectral image classification models, namely, LCFE-ViT, LCFE-SF, and LCFE-MF, are constructed based on recently proposed models. Extensive experiments demonstrate that the LCFE module effectively extracts local spectral signatures, and the proposed method enhances the performance of the various basic models. The code is available at https://github.com/YiSu1997/LCFE-X.

Comparative Analysis of Image Classification Methods on Cat Breeds and Behavior using Machine Learning Techniques

This research investigates image classification techniques applied to two distinct datasets related to cats. The primary focus is on addressing the problem of accurately classifying cat breeds and cat behavior. This research focuses on the comparative analysis of both deep learning and machine learning techniques. The techniques are categorized as use of Transfer learning on deep learning models, Transfer learning on machine learning algorithms, and Teachable machine pre-trained model. Transfer learning has gained popularity as one of the employed techniques, in inception V3 for classifying images. It requires re-utilizing an existing model, for a new model by applying a small-scale dataset to pace up training and enhance overall performance. Five different methodologies are explored: Convolutional Neural Networks using Google Inception-V3 model, Convolutional Neural Networks on top of Google Inception-V3 model with K-fold cross-validation, Random Forest on Inception V3 features, Support Vector Machine on Inception V3 features, and Teachable Machine model. The study aims to compare the performance of these methodologies in terms of accuracy, F1 score, and ROC-AUC score. The research results show that each approach has levels of effectiveness with various algorithms and models showing accuracy and F1 scores in classifying both cat breeds and behaviors. These findings offer information, on image classification, in datasets related to cats helping to improve the precision of identifying cat breeds and behaviors.

Lychee cultivar fine-grained image classification method based on improved ResNet-34 residual network

Lychee, a key economic crop in southern China, has numerous similar-looking varieties. Classifying these can aid farmers in understanding each variety's growth and market demand, enhancing agricultural efficiency. However, existing classification techniques are subjective, complex, and costly. This paper proposes a lychee classification method using an improved ResNet-34 residual network for six common varieties. We enhance the CBAM attention mechanism by replacing the large receptive field in the SAM module with a smaller one. Attention mechanisms are added at key network stages, focusing on crucial image information. Transfer learning is employed to apply ImageNet-trained model weights to this task. Test set evaluations demonstrate that our improved ResNet-34 network surpasses the original, achieving a recognition accuracy of 95.8442%, a 5.58 percentage point improvement.

An improved efficientnet B7 underwater marine organism image classification model based on the spatial group enhancement attention mechanisms

In this paper, we adopt the Efficientnet B7 model, which is an excellent performer in deep learning algorithms, and improve it by combining with the Spatial Group Enhancement attention mechanism to provide an innovative method for underwater biological image classification. After a reasonable division of the dataset, we introduced the improved Efficientnet B7 model for training. By observing the change of the loss curve on the training set, it can be seen that the loss value of the model gradually decreases from 0.6 to 0.05 and tends to be stable, which indicates that the classification prediction of the model is gradually close to the real result, and it can classify the underwater marine organisms images effectively. The accuracy on the training set gradually increases from the initial 80.12% to 98.31%, and the recall also grows from 75.82% to 98.93%, showing that the model predicts well. The performance on the test set shows an accuracy of 84.86%, a recall of 82.40%, and an F1 value of 80.2%, indicating that the model has good generalisation ability. This means that our proposed improved methods and ideas have achieved significant results in the task of underwater biological image classification, providing a useful reference for research in related fields. By comparing the experimental results, it can be seen that the optimisation of the Efficientnet B7 model by introducing the Spatial Group Enhancement attention mechanism achieves better performance in the underwater marine organisms image classification task, which lays the foundation for the future application of deep learning in underwater environments. This study is of positive significance for the promotion of underwater biology research, marine protection and other fields, and provides new ideas and methods for carrying out more in-depth related research.

An efficient image classification and segmentation method for crime investigation applications

AbstractThe field of forensic science is experiencing significant growth, largely driven by the increasing integration of holographic and immersive technologies, along with their associated head-mounted displays. These immersive systems have become increasingly vital in resolving critical crimes as they facilitate communication, interaction, and collaboration. Given the sensitive nature of their work, crime investigators require substantial technical support. There is a pressing need for accurate documentation and archiving of crime scenes, which can be addressed by leveraging 3D scanned scenes to accurately represent evidence and expected scenarios. This study aims to develop an enhanced AR. system that can be deployed on hologram facilities such as the Microsoft HoloLens. The proposed system encompasses two main approaches, namely image classification and image segmentation. Image classification utilizes various deep learning models, including lightweight convolutional neural networks (CNNs) and convolutional Long-Short Term Memory (ConvLSTM). Additionally, the image segmentation approach is based on the fuzzy active contour model (FACM). The effectiveness of the proposed system was evaluated for both classification and segmentation tasks, utilizing metrics such as accuracy, sensitivity, precision, and F1 score. The simulation results indicate that the proposed system achieved a 99% accuracy rate in classification and segmentation tasks, positioning it as an effective solution for detecting bloodstain patterns in AR applications.

ALFREDO: Active Learning with FeatuRe disEntangelement and DOmain adaptation for medical image classification

State-of-the-art deep learning models often fail to generalize in the presence of distribution shifts between training (source) data and test (target) data. Domain adaptation methods are designed to address this issue using labeled samples (supervised domain adaptation) or unlabeled samples (unsupervised domain adaptation). Active learning is a method to select informative samples to obtain maximum performance from minimum annotations. Selecting informative target domain samples can improve model performance and robustness, and reduce data demands. This paper proposes a novel pipeline called ALFREDO (Active Learning with FeatuRe disEntangelement and DOmain adaptation) that performs active learning under domain shift. We propose a novel feature disentanglement approach to decompose image features into domain specific and task specific components. Domain specific components refer to those features that provide source specific information, e.g., scanners, vendors or hospitals. Task specific components are discriminative features for classification, segmentation or other tasks. Thereafter we define multiple novel cost functions that identify informative samples under domain shift. We test our proposed method for medical image classification using one histopathology dataset and two chest X-ray datasets. Experiments show our method achieves state-of-the-art results compared to other domain adaptation methods, as well as state of the art active domain adaptation methods.

Medical Image Classification using a Many to Many Relation, Multilayered Fuzzy Systems and AI

One of the research gaps in the medical sciences is the study of orphan diseases or rare diseases, due to limited data availability of rare diseases. Our previous study addressed this successfully by developing an Artificial Intelligence (AI)-based medical image classification method using a multilayer fuzzy approach (MFA), for detecting and classifying image abnormalities for large and very small datasets. A fuzzy system is an AI system used to handle imprecise data. There are more than three types of fuzziness in any image data set: 1) due to a projection of a 3D object on a 2D surface, 2) due to the digitalization of the scan, and 3) conversion of the digital image to grayscale, and more. Thus, this was referred to in the previous study as a multilayer fuzzy system, since fuzziness arises from multiple sources. The method used in MFA involves comparing normal images containing abnormalities with the same kind of image without abnormalities, yielding a similarity measure percentage that, when subtracted from a hundred, reveals the abnormality. However, relying on a single standard image in the MFA reduces efficiency, since images vary in contrast, lighting, and patient demographics, impacting similarity percentages. To mitigate this, the current study focused on developing a more robust medical image classification method than MFA, using a many-to-many relation and a multilayer fuzzy approach (MCM) that employs multiple diverse standard images to compare with the abnormal image. For each abnormal image, the average similarity was calculated across multiple normal images, addressing issues encountered with MFA, and enhancing versatility. In this study, an AI-based method of image analysis automation that utilizes fuzzy systems was applied to a cancer data set for the first time. MCM proved to be highly efficient in detecting the abnormality in all types of images and sample sizes and surpassed the gold standard, the convolutional neural network (CNN), in detecting the abnormality in images from a very small data set. Moreover, MCM detects and classifies abnormality without any training, validation, or testing steps for large and small data sets. Hence, MCM may be used to address one of the research gaps in medicine, which detects, quantifies, and classifies images related to rare diseases with small data sets. This has the potential to assist a physician with early detection, diagnosis, monitoring, and treatment planning of several diseases, especially rare diseases.

Feature Consistency-Based Prototype Network for Open-Set Hyperspectral Image Classification.

Hyperspectral image (HSI) classification methods have made great progress in recent years. However, most of these methods are rooted in the closed-set assumption that the class distribution in the training and testing stages is consistent, which cannot handle the unknown class in open-world scenes. In this work, we propose a feature consistency-based prototype network (FCPN) for open-set HSI classification, which is composed of three steps. First, a three-layer convolutional network is designed to extract the discriminative features, where a contrastive clustering module is introduced to enhance the discrimination. Then, the extracted features are used to construct a scalable prototype set. Finally, a prototype-guided open-set module (POSM) is proposed to identify the known samples and unknown samples. Extensive experiments reveal that our method achieves remarkable classification performance over other state-of-the-art classification techniques.

Extracting waterline and inverting tidal flats topography based on Sentinel-2 remote sensing images: A case study of the northern part of the North Jiangsu radial sand ridges

Muddy tidal flats with landforms vary widely over space and time. The evolution of tidal flats topography has had a significant impact on the region. Therefore, there is a need to develop methods for rapidly and accurately characterizing tidal flats topography. This study applied thresholding segmentation (TS), machine learning, and two image classification methods to Sentinel-2 remote sensing images to extract the instantaneous waterline of the northern part of the North Jiangsu Radial sand Ridges (NJRSR). Tidal level data at specific time intervals were obtained using the Delft3D and TPXO9 models, after which tidal flats topography inversion was conducted. As a result, varied among the different classification methods, with the K-Nearest Neighbors (KNN) algorithm producing the optimal waterline extraction. Furthermore, the accuracy of tidal level simulation was improved by employing Delft3D software and constructing a scaled-down model of the study area. Comparisons with Delft3D model simulations of tidal levels with observations at three tide gauge stations yielded accuracies of 17.5 cm, 18.5 cm and 17.9 cm, this result showed that the establishment of the tidal level model can enhance the accuracy of tidal flats Digital Elevation Model (DEM) inversion. In fact, the combined application of the KNN algorithm and Delft3D model within retrieval of tidal level provided the most accurate topographic inversion, with an average error of 0.297 m, which was reduced to 0.292 m after median filtering, and the results indicated a trend of southward erosion in the tidal flats in the study area, with a turning point roughly located in the southern part of Doulong Port. The results of this study can help improve the future monitoring of dynamic changes in intertidal wetland topography and the conservation and development of tidal flats areas.

Image Classification Using CNN with CIFAR-10 Dataset

Abstract: Image classification is a very important issue in digital image analysis. CNN is type of Deep Neural Networks (DNN) that contains multiple layerssuch as Conv layers, integration layer and fully integrated layer. Convolutional neural networks (CNNs) are widely used in pattern-and picture- recognition problems as they have a number of advantages compared to others strategies. CIFAR-10is a very popular computer vision database. Thisdatabase is well read in it many types of in-depth study of object recognition. This database contains 60,000 images separated by 10 target classes, each a section containing 6000 images of 32 * 32 shapes. This database contains images of low-resolution (32 * 32), which allows researchers to experiment with new algorithms. Image classification is a fundamentaltask in computer vision that involves assigning a labelor category to an image. Convolutional Neural Networks (CNNs) have become the state-of-the-art method for image classification due to their ability toautomatically learn features from raw pixel values. Inthis project, we aim to build a CNN model that can accurately classify images in the CIFAR-10 dataset.

Classification of recurrent depression using brain CT images through feature fusion

ObjectiveThe objective of the study is to address the issues of high workload and low diagnostic efficiency in clinical medicine by employing computer-aided diagnostic analysis methods. The focus is on the identification of depressed patients using medical image data, particularly brain CT images. MethodsThe study proposes a CT image classification method for identifying depression disorder based on deep learning theory. Key methods include migration learning and feature fusion. Preprocessing and data enhancement techniques are utilized to filter out unnecessary features. An attention module is employed to better extract deep feature information. Additionally, a binary classification focus loss function is applied to address the unbalanced distribution of the dataset. ResultsExperimental results indicate that the proposed method achieves an image classification accuracy of 97.79%. Compared to a single model, there is an increase in accuracy by 2.61% and 1.81%, respectively. This improvement effectively enhances the classification accuracy of brain CT images with depressive disorders. The classification model also demonstrates better generalization performance. ConclusionThe study concludes that the proposed CT image classification method, incorporating migration learning, feature fusion, preprocessing, and attention modules, significantly improves the accuracy and classification speed of identifying depressive disorders in brain CT images. The model's generalization performance is highlighted, providing effective support for the enhancement of medical image classification accuracy.