<![CDATA[Iceberg lettuce was selected as the research object because its leaves show clear visual responses to nutrient deficiencies, while visual inspection methods commonly used by small- and medium-scale growers remain subjective. This study aims to develop an automated detection system for nutrient deficiency using an autoencoder combined with a Multilayer Perceptron (MLP). The dataset was sourced from Kaggle and categorized into four classes: Nitrogen, Phosphorus, Potassium, and Healthy. The preprocessing stage included converting images from BGR to HSV, resizing to 128×128 pixels, normalizing to 0–1, and applying an 80:20 train–test split. Feature extraction was performed using the autoencoder (encoder, bottleneck, decoder), while classification was carried out using the MLP (input, hidden, and output layers). Evaluation using a confusion matrix showed an accuracy of 86%, precision of 89%, recall of 87%, and an F1-score of 88%. The system has been implemented in a user-friendly web application that allows users to upload images and obtain detection results instantly. In conclusion, integrating autoencoder and MLP proved effective for automated nutrient deficiency detection in iceberg lettuce, providing a more objective alternative to conventional visual diagnosis.]]>
Copyrights © 2025
