Teaching Teachers First and Then Student: Hierarchical Distillation to Improve Long-Tailed Object Recognition in Aerial Images

Remote sensing data distribution generally exposes the long-tailed characteristic. This will limit the object recognition performance of existing deep models when they are trained with such unbalanced data. In this article, we propose a novel hierarchical distillation framework (HDF) to address the long-tailed object recognition in aerial images. First, we notice that not only student model should learn feature representations from teachers but also teacher models should learn feature representations from each other. Therefore, we build hierarchical teacher-wise distillation (HTWD) to improve the feature representations of the teacher models trained with middle and tail data, which is achieved by distilling the feature representations of the teacher model trained with head data. Second, we notice that the feature representations of the middle and tail classes cannot be effectively distilled from the teacher to the student since too little middle and tail data can be used to learn. Thus, we propose self-calibrated sampling (SCS) learning that enforces the student to strengthen the learning of the middle and tail data, thereby improving the student’ feature learning ability. Extensive experiments on two widely used DOTA and FGSC-23 datasets demonstrate the superior performance of the proposed method compared with state-of-the-art methods. Model and code are publicly available at https://github.com/wdzhao123/T2FTS .