Deep-Transfer Learning With Ensemble Features to Detect the Paddy Disease: a Study

Recently, a number of computerized methods are proposed to analyze the agriculture data to confirm the yield. Leaf image based plant condition monitoring is one of the common practices and to automate this process, a number of disease examination techniques are proposed and implemented. This research aims to propose a deep-transfer learning technique (DTLT) to examine the paddy plant disease based on the chosen digital images. To achieve a better detection, this work considered the ensemble of features (EoF) achieved with the chosen DL-models. The stages in this scheme includes, (i) data collection and resizing it to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$224 \times 224$</tex> pixels, (ii) feature extraction using the chosen model, (iii) best model selection based on the detection accuracy with SoftMaxclassifier, (iv) development of EoF using the selected model features and executing the detection using 3-fold crossvalidation. This work considered 800 images per class for the examination and the attained outcome confirms that the presented technique provides a detection accuracy of 100 % on the chosen model when EoF based detection is executed.