A Part-based Deep Learning Network for identifying individual crabs using abdomen images

Crabs, such as swimming crabs and mud crabs, are famous for their high nutritional value but difficult to preserve. Thus, the traceability of is vital food safety. Existing deep-learning methods can be applied identify individual crabs. However, there no previous study that used abdomen images In this paper, we provide a novel Part-based Deep Learning Network (PDN) reliably an crab from its captured under various conditions. our PDN, developed three non-overlapping overlapping partitions strategies image further designed part attention block. A (Crab-201) dataset with 201 more complex (Crab-146) were collected train test proposed PDN. Experimental results show PDN using partition strategy better than strategy. The edge texture has identifiable features sulciform lower abdomen. It also demonstrates PDN_OS3, which emphasizes strategies, reliable accurate counterpart crab.