Descriptor Distillation: Enhancing Local Descriptor Learning with DesDis Framework
The author proposes the DesDis framework to improve local descriptor learning by utilizing a teacher-student regularizer, resulting in better performances of student models compared to their teachers.