Hierarchical classification method for wildlife monitoring images based on species classification tree
To address the problem of high cost in the review of model results in species classification in wildlife monitoring images,this work proposes a hierarchical classification method based on the species classification tree.This method simultaneously performs zero-shot classification at the five classification levels of Class,Order,Family,Genus and Species in the species classification tree and reduces the labor cost of reviewing model results by providing more information about species identification.The proposed method utilizes hierarchical relationships between categories and introduces soft decision and path correction strategies.The coarse-grained results provided by the model are also valuable,while classification fails at the fine-grained species level.Compared to the baseline method enhanced by the path correction strategy,the accuracy of the proposed method increased by 0.27%,1.44%,1.42%,1.39%,and 1.03%,respectively,at each classification level,and the mistake severity was reduced by 1.1%.The proposed method improves the accuracy and consistency of hierarchical classification,enhances the interpretability of classification results,reduces the labor cost of reviewing model results,and increases ecologists' confidence and willingness to use deep learning models.
万方数据
维普
