In computer vision, the convolutional neural network (CNN) is a very popular model used for emotion recognition. It has been successfully applied to detect various objects in digital images with remarkable accuracy. In this paper, we extracted learned features from a pre-trained CNN and evaluated different machine learning (ML) algorithms to perform classification. Our research looks at the impact of replacing the standard SoftMax classifier with other ML algorithms by applying them to the FC6, FC7, and FC8 layers of Deep Convolutional Neural Networks (DCNNs). Experiments were conducted on two well-known CNN architectures, AlexNet and VGG-16, using a dataset of masked facial expressions (MLF-W-FER dataset). The results of our experiments demonstrate that Support Vector Machine (SVM) and Ensemble classifiers outperform the SoftMax classifier on both AlexNet and VGG-16 architectures. These algorithms were able to achieve improved accuracy of between 7% and 9% on each layer, suggesting that replacing the classifier in each layer of a DCNN with SVM or ensemble classifiers can be an efficient method for enhancing image classification performance. Overall, our research demonstrates the potential for combining the strengths of CNNs and other machine learning (ML) algorithms to achieve better results in emotion recognition tasks. By extracting learned features from pre-trained CNNs and applying a variety of classifiers, we provide a framework for investigating alternative methods to improve the accuracy of image classification.

Breast cancer is the second leading cause of mortality among women worldwide, highlighting the need for efficient histopathology-based screening methods for early diagnosis. This study introduces HistoDX, a deep learning framework to classify Invasive Ductal Carcinoma (IDC) using 277,524 histopathology image patches (50x50 pixels) from Paul Mooney’s IDC dataset on Kaggle, comprising No Cancer and IDC(+) classes. HistoDX employs a preprocessing pipeline with normalization, data augmentation, and class balancing via oversampling and weighted loss to address the class imbalance. A customized convolutional neural network, built on EfficientNetV2-B3 with additional layers, achieves 97% accuracy and a 0.91 ROC-AUC score on the test set. Validation on BreakHis (97% accuracy) and BACH (90% accuracy) datasets confirm generalizability, though detecting minority IDC(+) cases remains challenging. Low training and test losses underscore reliability. HistoDX empowers pathologists by enhancing diagnostic efficiency and minimizing subjectivity through effective class imbalance mitigation. Its robust performance across diverse datasets like BreakHis and BACH suggests readiness for clinical integration. Future research into advanced augmentation techniques, ensemble models, and whole-slide image analysis could further optimize accuracy, sensitivity, and scalability, paving the way for broader adoption in precision oncology.

Facial expression recognition (FER) is advancing human-computer interaction, especially, today, where facial masks are commonly worn due to the COVID-19 pandemic. Traditional unimodal techniques for facial expression recognition may be ineffective under these circumstances. To address this challenge, multimodal approaches that incorporate data from various modalities, such as voice expressions, have emerged as a promising solution. This paper proposed a novel multimodal methodology based on deep learning to recognize facial expressions under masked conditions effectively. The approach utilized two standard datasets, M-LFW-F and CREMA-D, to capture facial and vocal emotional expressions. A multimodal neural network was then trained using fusion techniques, outperforming conventional unimodal methods in facial expression recognition. Experimental evaluations demonstrated that the proposed approach achieved an accuracy of 79.81%, a significant improvement over the 68.81% accuracy attained by the unimodal technique. These results highlight the superior performance of the proposed approach in facial expression recognition under masked conditions.

Speech emotion recognition (SER) is an emerging technology that utilizes speech sounds to identify a speaker’s emotional state. Computational intelligence is receiving increasing attention from academics, health, and social media applications. This research was conducted to identify emotional states in verbal communication. We applied a publicly available dataset called RAVDEES. The data augmentation process involved adding noise, applying time stretching, shifting, and pitch, and extracting the features zero cross rate (ZCR), chroma shift, Mel-Frequency Cepstral Coefficients (MFCC), and a spectrogram. In addition, we used many pretrained deep learning models, such as VGG16, ResNet50, Xception, InceptionV3, and DenseNet121. Out of all of the deep learning models, Xception yielded superior outcomes. Furthermore, we improved performance by changing the Xception model to include hyperparameters and additional layers. We used a variety of performance evaluation parameters to test the proposed model. These included F1-score, accuracy, misclassification rate (MCR), precision, sensitivity, specificity, negative predictive value, false negative rate, false positive rate, false discovery rate, false omission rate, and false discovery rate. The model that we suggested demonstrated an overall accuracy of 98%, with an MCR of 2%. Additionally, it attained precision, sensitivity, and specificity values of 91.99%, 91.78%, and 98.68%, respectively. Additional models attained an F1-score of 91.83%. Our suggested model demonstrated superiority compared to other cutting-edge techniques

This study presents an Auto-Encoder Convolutional Neural Network (AECNNs) approach for anomaly detection in high-dimensional datasets. Unsupervised learning-based algorithms have a strong theoretical foundation and are widely used for anomaly detection in high-dimensional datasets, but some limitations significantly reduce their performance. This study proposes an algorithm to address these limitations. The proposed AECNN combines various convolutional layers, feature extraction, dimensionality reduction, and data preprocessing and was evaluated using accuracy, precision, recall, and F1-score. The performance of the proposed model was evaluated using a large real benchmark dataset. The proposed CNN-based autoencoder distinguished anomalies with an AUC score of 0.83 and remarkable accuracy, precision, recall, and F1 score.

This study focuses on Diabetic Retinopathy (DR), a disease caused by diabetes that affects the retina of the eye and eventually leads to blindness. Diabetes development progresses to retinopathy and must be addressed at an early stage for effective treatment. Currently, DR is classified as Non-Proliferative DR (NPDR) and Proliferative DR (PDR). This study proposes an Enhanced DR (P-EDR) method based on CNN using a high-resolution dataset benchmark of retinal images. Initially, the data were preprocessed by normalization, augmentation, and resizing to improve image quality and feature extraction. Evaluation was based on accuracy, specificity, sensitivity, and AUC-ROC. The proposed CNN-based P-EDR outperformed advanced ML strategies such as Support Vector Machine (SVM), Random Forest (RF), Probabilistic Neural network (PNN), and Gradient Boosting Machine (GBM) that were executed and compared to diagnose and classify DR. The proposed P-EDR extracts features such as a hemorrhage of the NPDR retina image to identify the disease using image processing for classification. P-EDR provides significant features from images in detection and classification, making it a successful model for diagnosing DR with improved accuracy of 93%, sensitivity of 92%, specificity of 94%, and AUC-ROC of 0.97%. These results highlight the potential of a P-EDR-based machine learning model to support ophthalmologists with the early and precise detection of DR, eventually helping with appropriate treatment and prevention of vision loss.