Autism Spectrum Disorder (ASD) is a neurodevelopmental condition that requires early diagnosis for effective intervention. Traditional diagnostic tools, such as MRI and CT scans, are often expensive, time-consuming, and inaccessible in underprivileged regions. To address this challenge, this study leverages facial images as a cost-effective and non-invasive alternative for ASD identification. A comprehensive evaluation of twelve pre-trained deep learning models—including ResNet-50, ResNet-101, ResNet-152, MobileNetV2, MobileNetV3, AlexNet, InceptionV1 (GoogleNet), SqueezeNet, EfficientNetB0, DenseNet121, DenseNet201, and VGG16—was conducted. Among these, DenseNet121 emerged as the top-performing model, achieving an accuracy of 90.33%, precision of 92.00%, recall of 92.00%, and an F1-score of 90.00%. Explainable AI techniques, including Local Interpretable Model-Agnostic Explanations (LIME) and Gradient-weighted Class Activation Mapping (Grad-Cam), were applied to highlight facial regions crucial for the model's predictions, enhancing transparency and trust. The proposed DenseNet121 model outperformed previous works. The results demonstrate the efficacy of this approach, offering a reliable, interpretable, and accessible solution for ASD identification, particularly in resource-constrained settings.

This research introduces “XMP‐Net,” a modified Xception–based deep learning architecture constructed for the categorization of skin conditions, with a particular focus on identifying monkeypox. The study recognizes skin images of four categories: normal, chickenpox, measles, and monkeypox. To enhance interpretability and foster confidence in the model′s predictions, Grad‐CAM (gradient‐weighted class activation mapping) and LIME (local interpretable model‐agnostic explanations) were employed to illustrate the model′s thinking manner. The model demonstrated impressive classification performance, attaining an accuracy of 98.33% for normal skin, 98.25% for monkeypox, 84.21% for measles, and 77.27% for chickenpox. Precision, recall, and F 1‐score values were also analyzed for each class, with monkeypox achieving a precision of 91.80%, a recall of 98.25%, and an F 1‐score of 94.92%. The visual explanations generated by Grad‐CAM and LIME highlighted critical parts in the input images that affected the model′s likelihoods, offering clinicians valuable insights into the diagnostic process. This research underscores the potential of explainable artificial intelligence (XAI) in augmenting traditional diagnostic methods, particularly for emerging communicable maladies like monkeypox, and provides a foundation for developing reliable, interpretable, and accessible diagnostic tools for resource‐constrained settings.

The significance of identifying human activity for the purpose of ensuring security is progressively growing day by day. Human activity recognition is an innovative technology that detects and recognizes human actions inside video footage. This research involves performing deep learning based YOLOv5s algorithms for the detection and recognition of four types of human activity, like sitting, standing, running, and sleeping. This research uses its own data set, which contains a total of 2375 images. There are 1745 images used for training and 630 images used for validation. We use Makesense. AI to classify the datasets. After classifying, we use the YOLOv5s method to train it. For this research, we utilized the Google Colabatory Environment with a V100 GPU. The results of this research indicate a precision score of 95.6%, a recall value of 95.8%, a f1-score of 95.69%, and an accuracy of 97%. The results of the experiment demonstrate that the model is capable of accurately identifying human activity. The model's capability to accurately identify human activity enables its utilization for monitoring purposes in various institutions such as daycare centers, schools, colleges, and hospitals.

Alzheimer’s disease (AD) is a degenerative brain disorder affecting millions worldwide, making early diagnosis crucial for effective management. Conventional diagnostic procedures, like neuropsychological exams and biomarker-based studies, are sometimes costly, intrusive, and not broadly accessible. This paper presents a strong deep learning (DL) architecture for classifying AD stages using MRI data. Our methodology employs transfer learning with pre-trained models such as InceptionV3, DenseNet169, and ResNet50V2, in conjunction with the synthetic minority over-sampling technique to mitigate class imbalance and improve model efficacy. We further use gradient-weighted class activation mapping to enhance the explainability and interpretability of predictions from the model, providing visual insights into the principal brain areas influencing classification judgments. Our suggested technique is assessed utilizing a publicly accessible dataset of MRI scans, yielding outstanding results in multiclass dementia stage classification, with an accuracy of 99.76%. The suggested approach surpasses current methodologies and offers a clear, medically interpretable approach for early Alzheimer’s diagnosis, which has possible applicability in resource-limited healthcare settings. These results underscore the efficacy of DL for non-invasive, affordable, and dependable AD diagnosis. However, external validation on datasets such as ADNI and OASIS is necessary to confirm generalizability before clinical application.

Detection, tracking, and re-identification (ReID) of children wearing similar uniforms in a kindergarten environment is a very complex challenge for computer vision. Traditional surveillance systems or simple convolutional neural network (CNN) models often fail to distinguish children in crowds and occlusions. To address this challenge, this study proposes a novel hybrid framework combining YOLOv8 and Vision Transformer (ViT). Using YOLOv8 for detection and ViT for global feature extraction, we trained the model on a custom dataset of 31,521 images, achieving an overall accuracy of 93.75%, and the public benchmark MOT20 dataset of 28,630 images, achieving an overall accuracy of 96.02%. Our system showed remarkable success in tracking performance, where it achieved 86.7% MOTA and 99.7% IDF1 scores. This high IDF1 score proves that the model is highly effective in preventing identity switch. The main novelty of this study is the behavioral analysis of children beyond the boundaries of surveillance, where we measure walking distance and trajectory, and screen time. Finally, through cross-dataset comparison with the MOT20 public benchmark, we demonstrated that our proposed customized model is much more effective than current state-of-the-art methods in overcoming the domain gap in specific environments such as kindergarten.

Autism Spectrum Disorder (ASD) is a neurodevelopmental condition that requires early diagnosis for effective intervention. Traditional diagnostic tools, such as MRI and CT scans, are often expensive, time-consuming, and inaccessible in underprivileged regions. To address this challenge, this study leverages facial images as a cost-effective and non-invasive alternative for ASD identification. A comprehensive evaluation of twelve pre-trained deep learning models—including ResNet-50, ResNet-101, ResNet-152, MobileNetV2, MobileNetV3, AlexNet, InceptionV1 (GoogleNet), SqueezeNet, EfficientNetB0, DenseNet121, DenseNet201, and VGG16—was conducted. Among these, DenseNet121 emerged as the top-performing model, achieving an accuracy of 90.33%, precision of 92.00%, recall of 92.00%, and an F1-score of 90.00%. Explainable AI techniques, including Local Interpretable Model-Agnostic Explanations (LIME) and Gradient-weighted Class Activation Mapping (Grad-Cam), were applied to highlight facial regions crucial for the model's predictions, enhancing transparency and trust. The proposed DenseNet121 model outperformed previous works. The results demonstrate the efficacy of this approach, offering a reliable, interpretable, and accessible solution for ASD identification, particularly in resource-constrained settings.