SHRIMP CLASSIFICATION VIA MULTI-VIEW FEATURE EXTRACTION USING PRETRAINED DEEP FEATURES

Shrimp classification is a complex task in computer vision applications. Very few research works have been carried out by researchers because shrimp features are difficult to extract due to their complex shape. The taste of shrimp-based foods depends on whether the same shrimp category is used. The export shrimp business relies on the quality of classification. Manual identification and classification of shrimps is a time-consuming process. In this work, Shrimps are classified based on different angular views, shape, and texture features. Features are extracted using both Zernike and pre-trained deep features. Further, shrimps are classified by regression. The qualitative and quantitative analysis shows that the proposed framework performs better with hand-crafted and deep features. The proposed framework with Zernike moments and VGG-16 features shows robustness with an accuracy of 89% and a precision of 91.7%.