An Improved Non-Destructive Shrimp Freshness Detection Method Based on Hybrid CNN and SVM with GAN Augmentation

In global seafood industries, production and consumption of Indian white shrimps (Vannamei) have considerably increased due to its sense of taste and nutritive value. At the same time, monitoring its perishable quality and shelf life to meet export standards is proving to be a major concern and also preserving the shrimps is very hard. The freshness of shrimp is one of the important qualities that consumers care about because deteriorated shrimp have many spoilage microorganisms which are harmful to human well-known. Deep learning technology exactly tackles the problem of selecting the features manually. However, the deep learning technology requires a large number of training samples. To address this, we adopted Generative Adversarial Network (GAN) for data augmentation to enhance the original training set. We proposed a hybrid CNN and SVM model with application of GAN based augmentation and investigated with traditional method which was also trained under different setting for extensive comparison study. The qualitative results show that model trained under GAN augmentation shows better identity. Lastly, we conclude that the hybrid CNN and SVM presents lower complexity with higher accuracy of 98.1%, suitable for real time and automated technique in industrial-level applications.