Automatic Labeling and Data Enhancement System for Sample Collection Based on Deep Learning
Jinghui Li, Chunge Li · 2022 6th International Conference on Intelligent Computing and Control Systems (ICICCS) · 2022
In this paper, sample collection based on deep learning is used to automatically label the problem, and the prior knowledge of the image label is used as the supervision information of the deep neural network. On the basis of obtaining the basic label vocabulary, the dependency relationship and prior distribution of the original image label vocabulary are used to improve the labeling result of the image. The advanced data augmentation method as the benchmark has shown unparalleled appeal at present. GAN learns the simulated distribution of images through game training and generates fake and real images. It also has broad prospects in the field of image migration and image editing.