Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{209781,
        author = {Mr. Rameshwar S. Devhade and Dr. Ranjana D. Raut},
        title = {Transfer learning based automated classification of cotton leaf diseases using field collected images from the western vidarbha region: A comparative study},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {5},
        pages = {3255-3261},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=209781},
        abstract = {Cotton is an economically important crop, but its production is significantly affected by various diseases and pests that can reduce crop quality and yield. Manual identification of cotton leaf diseases is time-consuming and depends on expert knowledge of the disease. This study presents an automated cotton leaf disease classification approach using transfer learning and field-collected images from the Western Vidarbha region of India. Initially, 9,361 cotton leaf images were collected using a mobile phone under field conditions.  A total of 9,361 cotton leaf images were initially collected using a mobile phone camera under agricultural-field conditions. After systematic duplicate checking and data cleaning, 9,180 unique images were selected. The dataset consisted of six classes: Aphids, Army_worm, Bacterial Blight, Healthy, Powdery Mildew, and Target spot. The cleaned dataset was divided using an 80:20 stratified split, resulting in 7,341 training images and 1,839 validation images. All images were resized to 224 × 224 pixels in size. Four transfer learning architectures, namely MobileNetV2, EfficientNetB0, ResNet50, and DenseNet121, were trained and evaluated using accuracy, precision, recall, F1-score, and confusion matrix analysis. Among the evaluated models, ResNet50 achieved the best performance, with an accuracy of 97.99%, followed by DenseNet121 (96.68%) and MobileNetV2 (94.24%). EfficientNetB0 achieved an accuracy of 23.55% under the present experimental configuration. The findings indicate that ResNet50 provides a promising foundation for the automated identification of cotton diseases. Future work will focus on improving model generalization using diverse field conditions and developing lightweight, explainable models for practical mobile-based applications.},
        keywords = {Cotton leaf disease, Western Vidarbha, Transfer learning, Deep learning, ResNet50, DenseNet121, MobileNetV2, EfficientNetB0, Field-collected images, Disease classification.},
        month = {October},
        }