Leaf Nutrient Deficiency Prediction using Deep Learning Models

Leaf nutrient deficiencies pose a significant threat to agricultural productivity and sustainability. Plants require essential nutrients for healthy growth, and their absence can lead to various symptoms, including color changes and leaf curling. Accurate diagnosis and timely detection are crucial to mitigate these issues in agricultural fields. This research study proposes a CNN-based approach to predict leaf nutrient deficiencies from images of leaves. This study employs three popular CNN architectures: InceptionV3, InceptionResNetV2, and AlexNet. The InceptionV3 model was specifically designed to predict copper, iron, magnesium, molybdenum, nitrogen, phosphorus, and potassium deficiencies. We utilized pre-trained ImageNet weights as a starting point and further fine-tuned the models with a dataset of leaf images. The InceptionV3 model achieved an accuracy of 97.8% on the training data, while InceptionResNetV2 and AlexNet reached accuracies of 97.7% and 92.22%, respectively. These results demonstrate the effectiveness of our proposed CNN-based approach in accurately diagnosing leaf nutrient deficiencies.