Formalin Detection in Fish Using EfficientNet-B3 and VGG-16

Fish is a main cuisine and a primary non-vegetarian food for the people of Bangladesh. Since Bangladesh is a riverine nation, fish can be found all over the place. Fish rots quickly, so the majority of fish must be preserved to meet the nutritional needs of the population. Sometimes, fish is imported and harvested from Bangladesh's rivers to feed the country's enormous population. Improper importers add formalin to their fish to keep it from spoiling during importation or international shipping. The effects of formalin on the human body are negative, causing a plethora of health issues, including cancer, miscarriages, tumors, etc. Therefore, it is crucial to identify fish that contain formalin and those that do not. Traditional methods of formalin detection rely on human senses, which can lead to incorrect results. Hence, a more dependable, automated method is required. Digital image processing has emerged as a potential tool for fish formalin detection in recent years. In this study, a complex application is proposed. Fish eyes can be used as the basis for an image processing system that distinguishes between formalin and non- formalin fish. One benefit of using digital image processing for formalin detection is that it offers a quick and discreet examination method. It can accurately and quickly detect formalin in fish photos without the need to prepare damaged samples. Digital image processing algorithms are more precise and consistent because they are not influenced by human biases or subjective assessments. The architectures we suggest are EfficientNET-B3 and VGG16. This system attained 100% training accuracy and dataset validation. However, formalin detection via digital image processing is not without challenges. The variability in fish appearances is one of the main obstacles that could prevent the development of a robust and efficient image processing program. This also includes the influence of additional factors, such as lighting. This paper focuses on testing an existing method proposed for this purpose.