An Enhanced Approach for Defect Detection: Implementing CNN Algorithm and Identifying Catastrophic Forgetting

Agriculture is a crucial sector that supports the global economy. With the increase of demands in agriculture, technology that supports this field keep on advancing. This research paper explores the variety of agricultural dataset to train the existing CNN model while testing the reliability of the model. The study expands the variety of fruit datasets that is suggested on the previous research. Some even suggest checking the reliability of the model used in defect detection. The proposed model is based on the previous study that has the best performance. The experiment is done thrice to check whether the model experienced catastrophic forgetting that is correlated with the reliability of the model. The result shows that the model can be trained to detect defect on a new variety of fruit dataset which is okra. It also shows no signs of catastrophic forgetting which means the model is reliable.