Finetuned Deep Learning Models for Fuel Classification: A Transfer Learning-Based Approach

Energies, Journal Year: 2025, Volume and Issue: 18(5), P. 1176 - 1176

Published: Feb. 27, 2025

Accurately classifying petrol and diesel fuel using an image processing method is crucial for fuel-related industries such as pumps, refineries, storage facilities. However, distinguishing between these fuels traditional methods can be challenging due to their similar visual characteristics. This study aims enhance the accuracy robustness of existing classification by utilizing transfer learning-based finetuned pre-trained deep learning models ensemble approaches. Specifically, we upgrade like ResNet152V2, InceptionResNetV2, EfficientNetB7 incorporating additional layers. Through learning, are adapted specific task fuels. To evaluate performance, upgraded model tested on a synthetic dataset. The results indicate that achieves recall, precision, f-score, scores 99.54%, 99.69%, 99.62%, 99.67%, respectively. Moreover, comparative analysis reveals outperform state-of-the-art baseline models.

Language: Английский

---

AI‐Powered Precision in Diagnosing Tomato Leaf Diseases

Complexity, Journal Year: 2025, Volume and Issue: 2025(1)

Published: Jan. 1, 2025

Correct detection of plant diseases is critical for enhancing crop yield and quality. Conventional methods, such as visual inspection microscopic analysis, are typically labor‐intensive, subjective, vulnerable to human error, making them infeasible extensive monitoring. In this study, we propose a novel technique detect tomato leaf effectively efficiently through pipeline four stages. First, image enhancement techniques deal with problems illumination noise recover the details clearly accurately possible. Subsequently, regions interest (ROIs), containing possible symptoms disease, captured. The ROIs then fed into K‐means clustering, which can separate sections based on health allowing diagnosis multiple diseases. After that, hybrid feature extraction approach taking advantage three methods proposed. A discrete wavelet transform (DWT) extracts hidden abstract textures in diseased zones by breaking down pixel values images various frequency ranges. Through spatial relation analysis pixels, gray level co‐occurrence matrix (GLCM) extremely valuable delivering texture patterns correlation specific ailments. Principal component (PCA) dimensionality reduction, selection, redundancy elimination. We collected 9014 samples from publicly available repositories; dataset allows us have diverse representative collection images. study addresses main diseases: curl virus, bacterial spot, late blight, Septoria spot. To rigorously evaluate model, split 70%, 10%, 20% training, validation, testing subsets, respectively. proposed was able achieve fantastic accuracy 99.97%, higher than current approaches. high precision achieved emphasizes promising implications incorporating DWT, PCA, GLCM, ANN an automated system diseases, offering powerful solution farmers managing efficiently.

Language: Английский

---

From Pixels to Diagnosis: Implementing and Evaluating a CNN Model for Tomato Leaf Disease Detection

Information, Journal Year: 2025, Volume and Issue: 16(3), P. 231 - 231

Published: March 16, 2025

The frequent emergence of multiple diseases in tomato plants poses a significant challenge to agriculture, requiring innovative solutions deal with this problem. paper explores the application machine learning (ML) technologies develop model capable identifying and classifying leaves. Our work involved implementation custom convolutional neural network (CNN) trained on diverse dataset leaf images. performance proposed CNN was evaluated compared against existing pre-trained models, i.e., VGG16 VGG19 which are extensively used for image classification tasks. further tested images leaves captured from real-world garden setting Greece. were carefully preprocessed an in-depth study conducted how either each preprocessing step or different—not supported by used—strain affects accuracy confidence detecting diseases.

Language: Английский

---

MoSViT: a lightweight vision transformer framework for efficient disease detection via precision attention mechanism

Frontiers in Artificial Intelligence, Journal Year: 2025, Volume and Issue: 8

Published: March 26, 2025

Maize, a globally essential staple crop, suffers significant yield losses due to diseases. Traditional diagnostic methods are often inefficient and subjective, posing challenges for timely accurate pest management. This study introduces MoSViT, an innovative classification model leveraging advanced machine learning computer vision technologies. Built on the MobileViT V2 framework, MoSViT integrates CLA focus mechanism, DRB module, Block, LeakyRelu6 activation function enhance feature extraction accuracy while reducing computational complexity. Trained dataset of 3,850 images encompassing Blight, Common Rust, Gray Leaf Spot, Healthy conditions, achieves exceptional performance, with accuracy, Precision, Recall, F1 Score 98.75%, 98.73%, 98.72%, respectively. These results surpass leading models such as Swin Transformer V2, DenseNet121, EfficientNet in both parameter efficiency. Additionally, model's interpretability is enhanced through heatmap analysis, providing insights into its decision-making process. Testing small sample datasets further demonstrates MoSViT's generalization capability potential small-sample detection scenarios.

Language: Английский

---

StrawberryNet: Fast and Precise Recognition of Strawberry Disease Based on Channel and Spatial Information Reconstruction

Agriculture, Journal Year: 2025, Volume and Issue: 15(7), P. 779 - 779

Published: April 3, 2025

Timely and effective identification diagnosis of strawberry diseases play essential roles in the prevention diseases. Nevertheless, various types with high similarity pose a great challenge to accuracy diseases, recent module parameter counts is not suitable for real-time monitoring. Therefore, this paper, we propose lightweight disease method, termed StrawberryNet, achieve accurate First, decrease number parameters, instead standard convolution, partial convolution selected construct backbone extracting features disease, which can significantly improve efficiency. And then, discriminative feature extractor, including channel information reconstruction network (CIR-Net) spatial (SIR-Net) modules, designed abstracting identifiable different disease. A large experimental results were conducted on constructed dataset, containing 2903 images 10 common normal leaves fruits. Extensive experiments show that recognition proposed method reach 99.01% only 3.6 M have good balance between precision speed compared other excellent modules.

Language: Английский

---