The development of RBL-STEM learning tools to improve students' computational thinking skills in solving plant disease classification problems using convolutional neural network segmentation

This research aims to develop Research Based Learning (RBL) and Science, Technology, Engineering and Mathematics (STEM) based learning tools to improve students' computational thinking skills. The tools include a student task design (RTM), a student worksheet (LKM) and a learning outcome test (THB). Using Thiagarajan's 4D development model (define, design, develop, disseminate), the device was tested on students using a Convolutional Neural Network (CNN) approach for citrus plant disease classification using data from quadcopter drones. The validation results showed that the device was valid with an average score of 3.85 (96.42%). The practicality of the device is very high with an implementation score of 3.85 (96.36%) and a positive student response of 90.08%. The effectiveness of the device was demonstrated by 90% of students achieving classical completeness on the post-test and an increase in computational thinking skills from 0% (high) on the pre-test to 92% on the post-test. The paired sample t-test results also confirmed the statistically significant increase. This learning tool proved to be valid, practical and effective, contributing to technology-based learning innovation in modern agriculture.