Research on Reliability Metrics for Deep Learning Visual Algorithms

Yuanxin Li, Guang Yang, Pengqi Wang, Zhili Zhang, Lingzhong Meng

Abstract: The reliability of deep learning visual algorithms, as the main source of perceptual information in many important applications, is of great importance. This article focused on the reliability metrics of deep learning visual algorithms, collecting and organizing domestic and foreign papers on deep learning algorithm measurement, public research reports, existing standards and norms at home and abroad. Mainly focusing on three typical application scenarios: object classification, object detection and object tracking, we analyzed the characteristics of various visual tasks, collected and summarized deep learning metrics for each type of algorithms, and established a set of deep learning visual algorithm metrics for different visual tasks. Finally, we chose some typical algorithms and datasets for experiments, then used these metrics to evaluate the algorithms and analyze the results.

External IDs:dblp:conf/isssr/LiYWZM24