Random Part Localization Model for Fine Grained Image Classification
Xin Qi, Tiejun Lv, Hui Gao · 2019
Fine-grained recognition is challenging due to its subtle local inter-class differences versus large intra-class variations. Finding those subtle traits that fully characterize the object is not straightforward. In this paper, we present a novel random part localization model, which first extracts the foreground object using the saliency map, and then localizes the discriminative parts through a set of potential regions in a random way based on their contribution to classification. We train three convolutional neural networks to capture the features that belong to different levels and average their classification results as our final prediction score. Experiments show that our approach achieves competitive performance compared with state-of-the-art methods on three publicly available fine-grained recognition datasets (CUB200-2011, Stanford Cars and FGVC-Aircraft).