Dataset Culling: Towards Efficient Training of Distillation-Based Domain Specific Models | IEEE Conference Publication | IEEE Xplore