Search Results (461)

This study examines the utilization of deep learning and transfer learning models for classifying photos of Indian cuisine. Indian cuisine, characterized by its extensive diversity and intricate presentation, poses considerable hurdles in food recognition owing to changes in ingredients, texture, and visual aesthetics. To tackle these challenges, we utilized a bespoke Convolutional Neural Network (CNN) and harnessed cutting-edge transfer learning models such as DenseNet121, InceptionV3, MobileNet, VGG16, and Xception. The research employed a varied dataset comprising 13 food categories and executed preprocessing techniques like HSV conversion, noise reduction, and edge identification to improve image quality. Metrics for performance evaluation, including accuracy, precision, recall, and F1-score, were employed to assess model efficacy. The CNN model demonstrated a mediocre performance, revealing overfitting concerns due to a substantial disparity between training and validation accuracy. In contrast, transfer learning models, particularly DenseNet121, InceptionV3, and Xception, exhibited an enhanced generalization ability, each attaining above 92% accuracy across all criteria. MobileNet and VGG16 produced reliable outcomes with marginally reduced performances. The results highlight the efficacy of transfer learning in food image classification and indicate that fine-tuned, pre-trained models markedly improve classification accuracy. This research advances the creation of intelligent food recognition systems applicable in dietary monitoring, nutrition tracking, and health management. Full article

Reinforcement Learning (RL) enables learning optimal decision-making strategies by maximizing cumulative rewards. Deep reinforcement learning (DRL) enhances this process by integrating deep neural networks (DNNs) for effective feature extraction from high-dimensional input data. Unlike prior studies focusing on algorithm design, we investigated the impact of different feature extractors, DNNs, on DRL performance. We propose an enhanced feature extraction model to improve control effectiveness based on the proximal policy optimization (PPO) framework in autonomous driving scenarios. Through a comparative analysis of well-known convolutional neural networks (CNNs), MobileNet, SqueezeNet, and ResNet, the experimental results demonstrate that our model achieves higher cumulative rewards and better control stability, providing valuable insights for DRL applications in autonomous systems. Full article

Coal and rock recognition is one of the key technologies in mining production, but traditional methods have limitations such as single-feature representation dimension, insufficient robustness, and unbalanced performance in lightweight design under noise interference and complex feature conditions. To address these issues, an Efficient Channel Attention Reparameterized Network (ECA-RepNet) based on recurrence plot and Efficient Channel Attention mechanism is proposed. The one-dimensional vibration signal is mapped to the two-dimensional image space through a recurrence plot (RP), which retains the dynamic characteristics of the time series while capturing the complex patterns in the signal. Multi-scale feature extraction and lightweight design are achieved through the reparameterized large kernel block (RepLK Block) and the depthwise separable convolution (DSConv) module. The ECA module is introduced to embed multiple convolutional layers. Through global average pooling, one-dimensional convolution, and dynamic weight allocation, the modeling ability of inter-channel dependencies is enhanced, the model robustness is improved, and the computational overhead is reduced. Experimental results demonstrate that the ECA-RepNet model achieves 97.33% accuracy, outperforming classic models including ResNet, CNN, and MobileNet in parameter efficiency, training time, and inference speed. Full article

Background: Fungal infections are considered a global health concern, resulting in high morbidity and mortality rates, especially in immunocompromised individuals. Traditional diagnostic techniques, such as microscopy, culture, and polymerase chain reaction (PCR), suffer from low sensitivity, long processing time, and accessibility challenges, especially in resource-limited settings. Artificial intelligence (AI) and machine learning (ML) tools have demonstrated potential to enhance diagnostic accuracy and efficiency. This systematic study assesses the progress, precision, and efficacy of AI-driven diagnostic tools for fungal infections within various clinical contexts in comparison to traditional procedures. Methods: A systematic review was conducted according to PRISMA principles. Literature searches were conducted in PubMed, ScienceDirect, Web of Science, and Ovid, focusing on research employing AI models to diagnose fungal infections. The inclusion criteria were research that compared AI-based tools with conventional diagnostic methods in terms of sensitivity, specificity, and accuracy. Data extraction and quality evaluation were performed utilizing validated instruments, such as the Methodological Index for Non-Randomized Studies (MINORS). Results: Eleven research studies met the inclusion criteria: six retrospective and five prospective investigations. AI models, such as convolutional neural networks (CNNs), Faster R-CNN, VGG19, and MobileNet, have improved diagnostic accuracy, sensitivity, and specificity compared to traditional methods. However, differences in dataset quality, model validation, and real-world applicability remain as limitations. Conclusions: AI-driven diagnostic technologies provide significant benefits in identifying fungal infections, improving the speed and accuracy of diagnoses. However, additional extensive investigations and clinical validation are required to improve model generalizability and facilitate smooth incorporation into healthcare systems. Full article

Rice (Oryza sativa L.) is a staple food for over half of the global population, with significant economic, agricultural, and cultural importance, particularly in Asia. Thousands of rice varieties exist worldwide, differing in size, shape, color, and texture, making accurate classification essential for quality control, breeding programs, and authenticity verification in trade and research. Traditional manual identification of rice varieties is time-consuming, error-prone, and heavily reliant on expert knowledge. Deep learning provides an efficient alternative by automatically extracting discriminative features from rice grain images for precise classification. While prior studies have primarily employed deep learning models such as CNN, VGG, InceptionV3, MobileNet, and DenseNet201, transformer-based models remain underexplored for rice variety classification. This study addresses this gap by applying two deep learning models such as Swin Transformer and Vision Transformer for multi-class classification of rice varieties using the publicly available PRBD dataset from Bangladesh. Experimental results demonstrate that the ViT model achieved an accuracy of 99.86% with precision, recall, and F1-score all at 0.9986, while the Swin Transformer model obtained an accuracy of 99.44% with a precision of 0.9944, recall of 0.9944, and F1-score of 0.9943. These results highlight the effectiveness of transformer-based models for high-accuracy rice variety classification. Full article

This study presents a robust deep learning-based approach for temperature sensing using Fiber Specklegram Sensors (FSS), leveraging an extended experimental framework to evaluate model generalization. A convolutional neural network (CNN), specifically a customized MobileNet architecture (MNet-reg), was trained on multiple experimental datasets to assess the impact of increasing data availability on sensing accuracy. Generalization is evaluated as cross-dataset performance under unseen experimental realizations, rather than under controlled intra-dataset splits. The experimental setup utilized a multi-mode optical fiber (MMF) (core diameter 62.5 µm) subjected to controlled thermal cycles via a PID-regulated heating system. The curated dataset comprises 24,528 specklegram images captured over a temperature range of 25.00 °C to 200.00 °C with increments of ~0.20 °C. The experimental results demonstrate that models trained with an increasing number of datasets (from 1 to 13) significantly improve accuracy, reducing Mean Absolute Error (MAE) from 13.39 to 0.69 °C, and achieving a Root Mean Square Error (RMSE) of 0.90 °C with an R² score of 0.99. Our systematic analysis establishes that scaling experimental data diversity—through training on multiple independent realizations—is the foundational strategy to overcome domain shift and enable robust cross-dataset generalization. Full article

The increasing size of recent deep neural networks, particularly when applied to learning across multiple domains, limits their deployment in resource-constrained environments. To address this issue, this study proposes a lightweight neural architecture with a parallel structure of convolutional layers to enable efficient and scalable multi-domain learning. The proposed network includes an individual feature extractor for domain-specific features and a common feature extractor for the shared features. This design minimizes redundancy and significantly reduces the number of parameters while preserving classification performance. To evaluate the proposed method, experiments were conducted using four image classification datasets: MNIST, FMNIST, CIFAR10, and SVHN. These experiments focused on classification settings where each image contained a single dominant object without relying on large pretrained models. The proposed model achieved high accuracy while significantly reducing the number of parameters. It required only 3.9 M parameters for learning across the four datasets, compared to 33.6 M for VGG16. The model achieved an accuracy of 98.87% on MNIST and 85.83% on SVHN, outperforming other lightweight models, including MobileNet v2 and EfficientNet v2b0, and was comparable to ResNet50. These findings indicate that the proposed architecture has the potential to support multi-domain learning while minimizing model complexity. This approach may be beneficial for applications in resource-constrained environments. Full article

With the rapid advancement of artificial intelligence (AI) technology, training deep neural networks has become a core computational task that consumes significant energy in data centers. Researchers often employ various methods to estimate the energy usage of data center clusters or servers to enhance energy management and conservation efforts. However, accurately predicting the energy consumption and carbon footprint of a specific AI task throughout its entire lifecycle before execution remains challenging. In this paper, we explore the energy consumption characteristics of AI model training tasks and propose a simple yet effective method for predicting neural network training energy consumption. This approach leverages training task metadata and applies genetic programming-based symbolic regression to forecast energy consumption prior to executing training tasks, distinguishing it from time series forecasting of data center energy consumption. We have developed an AI training energy consumption environment using the A800 GPU and models from the ResNet{18, 34, 50, 101}, VGG16, MobileNet, ViT, and BERT families to collect data for experimentation and analysis. The experimental analysis of energy consumption reveals that the consumption curve exhibits waveform characteristics resembling square waves, with distinct peaks and valleys. The prediction experiments demonstrate that the proposed method performs well, achieving mean relative errors (MRE) of 2.67% for valley energy, 8.42% for valley duration, 5.16% for peak power, and 3.64% for peak duration. Our findings indicate that, within a specific data center, the energy consumption of AI training tasks follows a predictable pattern. Furthermore, our proposed method enables accurate prediction and calculation of power load before model training begins, without requiring extensive historical energy consumption data. This capability facilitates optimized energy-saving scheduling in data centers in advance, thereby advancing the vision of green AI. Full article

Effective and transparent medical diagnosis relies on accurate and interpretable classification of medical images across multiple modalities. This paper introduces an explainable multi-modal image analysis framework based on a dual-stream architecture that fuses handcrafted descriptors with deep features extracted from a custom MobileNet. Handcrafted descriptors include frequency-domain and texture features, while deep features are summarized using 26 statistical metrics to enhance interpretability. In the fusion stage, complementary features are combined at both the feature and decision levels. Decision-level integration combines calibrated soft voting, weighted voting, and stacking ensembles with optimized classifiers, including decision trees, random forests, gradient boosting, and logistic regression. To further refine performance, a hybrid class-specific feature selection strategy is proposed, combining mutual information, recursive elimination, and random forest importance to select the most discriminative features for each class. This hybrid selection approach eliminates redundancy, improves computational efficiency, and ensures robust classification. Explainability is provided through Local Interpretable Model-Agnostic Explanations, which offer transparent details about the ensemble model’s predictions and link influential handcrafted features to clinically meaningful image characteristics. The framework is validated on three benchmark datasets, i.e., BTTypes (brain MRI), Ultrasound Breast Images, and ACRIMA Retinal Fundus Images, demonstrating generalizability across modalities (MRI, ultrasound, retinal fundus) and disease categories (brain tumor, breast cancer, glaucoma). Full article

The green-making stage is crucial for forming the distinctive aroma and flavor of Tieguanyin tea. Current green-making stage recognition relies on tea makers’ sensory experience, which is labor-intensive and time-consuming. To address these issues, this paper proposes a lightweight automatic recognition model named T-GSR for the accurate and objective identification of Tieguanyin tea green-making stages. First, an extensive set of Tieguanyin tea images at different green-making stages was collected. Subsequently, preprocessing techniques, i.e., multi-color-space fusion and morphological filtering, were applied to enhance the representation of target tea features. Furthermore, three targeted improvements were implemented based on the MobileNet V3 backbone network: (1) an adaptive residual branch was introduced to strengthen feature propagation; (2) the Rectified Linear Unit (ReLU) activation function was replaced with the Gaussian Error Linear Unit (GELU) to improve gradient propagation efficiency; and (3) an Improved Coordinate Attention (ICA) mechanism was adopted to replace the original Squeeze-and-Excitation (SE) module, enabling more accurate capture of complex tea features. Experimental results demonstrate that the T-GSR model outperforms the original MobileNet V3 in both classification performance and model complexity, achieving a recognition accuracy of 93.38%, an F1-score of 93.33%, with only 3.025 M parameters and 0.242 G FLOPs. The proposed model offers an effective solution for the intelligent recognition of Tieguanyin tea green-making stages, facilitating online monitoring and supporting automated tea production. Full article

Small-scale farms produce more than one-third of the world’s food supply, making them a crucial contributor to global food security. In this study, an artificial intelligence of things (AIoT) framework is introduced for smart small-scale farm management. For plant disease detection, the lightweight MobileViT model, which integrates vision transformer and convolutional modules, was utilized to efficiently capture both global and local image features. Data augmentation and transfer learning were employed to enhance the model’s overall performance. MobileViT resulted in a test accuracy of 99.5%, with per-class precision, recall, and f1-score ranging between 0.92 and 1.00 considering the benchmark Plant Village dataset (14 species–38 classes). MobileViT was shown to outperform several standard deep convolutional networks, including MobileNet, ResNet and Inception, by 2–12%. Additionally, an LLM-powered interactive chatbot was integrated to provide farmers with instant plant care suggestions. For plant environment management, the powerful, cost-effective ESP32 microcontroller was utilized as the core processing unit responsible for collecting sensor data (e.g., soil moisture), controlling actuators (e.g., water pump for irrigation), and maintaining connectivity with Google Firebase Cloud. Finally, a mobile application was developed to integrate the AI and IoT system capabilities, hence providing users with a reliable platform for smart plant disease detection and environment management. Each system component was each tested individually, before being incorporated into the mobile application and tested in real-world scenarios. The presented AIoT-based solution has the potential to enhance crop productivity within small-scale farms while promoting sustainable farming practices and efficient resource management. Full article

In snowy regions, road surface conditions change due to snowfall or ice formation in winter. This can lead to very dangerous situations when driving a car. Therefore, recognizing road surface conditions is important for both drivers and road managers. Road surface classification using in-vehicle cameras faces challenges due to the diverse environments in which vehicles operate. It is difficult to build a single classification model that can handle all conditions. One major challenge is illumination. During dusk, it changes rapidly and drastically, resulting in poor classification accuracy. Therefore, a robust method is needed to accurately recognize road conditions at all times. In this study, we used an image translation method to standardize illumination conditions. Next, we extracted features from both the translated images and the original images using MobileNet. Finally, we integrated these features using Late Fusion with an Extreme Learning Machine to classify road conditions. The effectiveness of this method was verified using a dataset of in-vehicle camera images. The results showed that the accuracy of this method achieved 78% during dusk and outperformed the comparison methods. It was confirmed that the uniformity of illumination conditions contributed to the improvement in classification accuracy. The proposed method can classify road conditions even during dusk, when sudden changes in illumination occur. This demonstrates the potential to realize a robust road condition recognition method that contributes to improved driver safety and efficient road management. Full article

This study aims to predict the freshness of sea bass (Dicentrarchus labrax) using deep learning models based on image data. For this purpose, 10 fish were monitored daily from the day of purchase until three days after spoilage, with multi-angle imaging (eight distinct perspectives per fish, both with and without background) and corresponding quality analyses. A total of 22 quality parameters—10 categorical (sensory-based) and 12 numerical (color-based)—were evaluated, with the purchasability parameter defined as the most critical indicator of freshness. Using seven popular transfer learning algorithms (EfficientNetB0, ResNet50, DenseNet121, VGG16, InceptionV3, MobileNet, and VGG19), 2464 predictive models (1120 classification and 1344 regression) were trained. Classification models were evaluated using accuracy, precision, recall, F1-score, and response time, while regression models were assessed using mean absolute error and tolerance-based error metrics. The results showed that the MobileNet algorithm achieved the best overall performance, successfully predicting 15 of the 22 parameters with the lowest error or highest accuracy. Importantly, in the prediction of the most critical parameter—purchasability—the DenseNet121 architecture yielded the best classification performance with an accuracy of 0.9894. The findings indicate that deep learning-based image analysis is a viable method for evaluating the freshness of fish. Full article

In order to maintain pedestrian safety in vehicle collisions and enhance collision safety, this paper proposes a rapid prediction method of head injuries for pedestrian protection based on deep learning, which could be used to design and optimize pedestrian protection performance during the vehicle design stage. However, traditional finite element simulation methods involve a large computational effort and long calculation time, and multiple computations are required to obtain the corresponding pedestrian head injury results for engine hood structural optimization. Therefore, to accelerate the design process and save time costs, this paper proposes a deep learning-based method for the rapid prediction of pedestrian head injuries. Compared with traditional finite element simulation techniques, this method will greatly improve the efficiency of obtaining head injury results, and its core lies in establishing a prediction model for pedestrian head injury results. The sample data for establishing the prediction model is defined initially, in which the head injury criterion (HIC) and vehicle structure serve as the output and input of the prediction model, respectively. The voxelization method is used to digitally express the car body structure. Convolutional neural networks (CNNs) such as ResNet50, MobileNet, SqueezeNet, and ShuffleNet are used to train the model. After adjusting the dataset and model hyperparameters, the prediction model with the smallest error is obtained. The cross-validation method was used to verify the robustness and generalization ability of the model. The average error rate of the obtained prediction model for predicting head injuries was 14.28%, which proved the accuracy and applicability of the prediction model. Full article

Rice lodging, a common physiological issue that occurs during rice growth and development, is a major factor contributing to a decline in rice production. Current techniques for the extraction of rice lodging are subjective and require considerable time and labor. In this paper, we propose a fully automated method in an end-to-end format to objectively calculate the rice-lodging rate based on remote sensing data captured by a drone under field conditions. An image post-processing method was applied to enhance the semantic-segmentation results of an operable lightweight model on an embedded board. The area of interest within the parcel was preserved based on these results, and the lodging occurrence rate was calculated in a fully automated manner without external intervention. Five models were compared based on the U-Net and lite-reduced atrous spatial pyramid pooling (LR-ASPP) models with MobileNet versions 1–3 as the backbones. The final model, MobileNetV1_U-Net, performed the best with an RMSE of 11.75 and R² of 0.875, and MobileNetV3_LR-ASPP (small) achieved the shortest processing time of 4.9844 s. This study provides an effective method for monitoring large-scale rice lodging, accurate extraction of areas of interest, and calculating lodging occurrence rates. Full article