A TEDE Algorithm Studies the Effect of Dataset Grouping on Supervised Learning Accuracy

Datasets are the basis for research on deep learning methods in computer vision. The impact of percentage training sets a dataset performance neural network models needs to be further explored. In this paper, twice equal difference enumeration (TEDE) algorithm is proposed investigate effect different set percentages model, and optimal determined. By selecting Pascal VOC dividing it into six datasets from largest smallest, then each analyzed according five percentages, YOLOv5 convolutional used train test 30 determine model corresponding percentages. Finally, tests were conducted using Udacity Self-Driving with self-made Tire Tread Defects (TTD) dataset. results show that superior when accounts between 85% 90% overall partitioning obtained by TEDE can provide reference research.