Few-Shot Learning for Vehicle Footprint Recognition: State-of-the-Art

Vehicle footprint recognition is an important means for obtaining clues in traffic accident control and criminal case solving. However, in practical applications, due to the limited number of available training samples, overfitting often occurs in the recognition task. Few-shot learning can effectively solve the problem of inaccurate recognition due to the scarcity of training data. This paper summarizes the state-of-the-art methods for few-shot vehicle footprint recognition and divides them into three categories including model fine-tuning-based, feature representation-based and feature distribution-based according to different learning mechanisms applied. In addition, experimental results are provided to compare the performance of representative methods in the three categories. Finally, future research directions in this field are discussed.