Poultry Meat Classification Using MobileNetV2 Pretrained Model

Abstract

In Morocco the meat business risks being targeted by fraud and adulteration, leading customers to probe the authenticity of the meat.The traditional styles for verifying meat types are expensive and consuming time.In this work, we propose a method based on computer vision and deep learning, which allows the bracket and isolation between turkey and chicken and Fayoumi and chicken farmer meat.We created a model grounded on the pre-trained Mobile Net V2 model and trained it with a Dataset containing the collected images of the four poultries.The evaluation of this model has given satisfactory results and has demonstrated that the model is suitable to predict the meat class with a delicacy of over 98%.The algorithm can be generalized to separate between authentic and fake meat.