Detection of Freshness of Fish using Machine Learning Techniques on Vyas Municipality, Nepal

Abstract

The historical narrative of the fish trade is well-document in various sources. However, the concerning prevalence of fish traders vending spoiled fish poses a significant threat to human health, prompting specific research inquiries. The study aimed to address key questions: What quality of healthy fish do traders sell? How effective are their fish storage methods? What's the duration between fish purchase and consumer access? The study objectives were devised to uncover a actual condition of the fish on sale, assess storage practices, and determine the selling timeline. To achieve these aims, the study employed the EfficientNetB1 machine learning model, chosen for its simplicity and high accuracy. Five fish shops and traders from wards 1,2,3 and 4 in Damauli, the primary city of Vyas Municipality in Nepal, were selected for investigation. Results from five main city shops in Damauli revealed that only 26% of the fish were deemed healthy, while a concerning 74% were identified as rotten. Similarly, within the sample, 44% of the fish were healthy, while 56% were spoiled. This study unveiled that fish were being sold even up to 15 days post-purchase, employing ice packs, refrigeration, and potentially chemicals for storage. These findings highlight the urgent need for ongoing monitoring by relevant stakeholders and local government entities to address this issue effectively.