Abstract:
Ethiopia is among the top 10 avocado-growing countries. Nowadays, avocados are becoming increasingly popular among producers and consumers in Ethiopia. Recently as reported by Ethiopia's Agricultural Transformation Agency (ATA), nine agricultural clusters in the Amhara region, North Mecha Woreda of West Gojjam Zone, have already started exporting Hass (dark green and bumpy type) avocados cultivated on more than 200 hectares of land. In Ethiopia, avocado production is constrained by many factors including poor quality and lack of appropriate postharvest technology that does not satisfy the international and domestic market demand in volume and quality. Studies on avocado production predict that Ethiopia will take advantage of the global avocado market by giving special attention to its potential and modernizing the private sector, small enterprises, service providers, and research and development. In doing so special attention has to be given to the postharvest handling practices to meet the quality standard of the international market. To address this problem, we proposed a Deep learning model that automatically classifies the avocado fruit based on their quality level. We collected the necessary dataset from the Fruit Cultivation Center based on their quality level. We used image processing techniques that include image preprocessing, segmentation, feature extraction, and prediction. We compared the newly proposed Avocado fruit quality grading using the MobileNetV2 model with other state-of-the-art models such as VGGNet, AlexNet, and Resenet50. The proposed model achieved 98.2% testing accuracy; AlexNet model achieved 86% testing accuracy; the ReseNet50 91% and the VGGNet model achieved 90% testing accuracy. So the proposed MobileNetV2 model for Avocado fruit quality grading outperforms the other models for quality grading. Keywords: MobileNetV2, feature Extraction, Avocado fruit, Image processing