Rescaling large datasets based on validation outcomes of a pre-trained network

Thanh Tuan Nguyen, Thanh Phuong Nguyen

Abstract: Highlights•Rescaling a large dataset based on the validation outcomes of a pre-trained network.•A challenging small subset is constructed by pointing out the sensitive images.•An application for rescaling ImageNet and Places365 to obtain their rescaled subsets.•Results have verified the challenge of rescaled subsets versus the original datasets•A network draft performs well on large datasets if it is good on the rescaled subsets.