Deep Learning Approach for Identifying Fresh and Rotten Chili Peppers using YOLO

Abstract

Abstract—In Indonesia, chili peppers are a vital agricultural commodity, widely used in various local dishes. Chili peppers are widely sold in modern markets and traditional markets. In some traditional markets, sellers sometimes do not store them in suitable conditions, thereby affecting the quality of the chili peppers. Sometimes, buyers are unable to select the specific chilies being sold wholesale, so they often end up with chilies of inferior quality. Therefore, this study proposes a deep learning-based computer vision system using the You Only Look Once (YOLO) framework to distinguish between fresh and rotten chili peppers. The model that we use was trained using 400 datasets from web-crawled images. To simulate real-world scenarios, during testing, some synthetic images were arranged in both regular and random patterns, as well as various chili orientations. Experimental results demonstrate that the proposed model effectively localizes and classifies chili peppers with confidence scores ranging from 0.55 to 0.90. Quantitative evaluation results achieved an accuracy of 0.5048, precision of 0.5516, recall of 0.4580, and [email protected] of 0.4662, indicating moderate detection and classification performance under varying visual conditions. The result of the experiment shows that the system success for distinguishes between fresh and rotten chilli pepper in some conditions of image variation, such as the angular orientation of the chilli peppers and their proximity. This shows that the approach using YOLO provides very promising results for application in the agricultural market.