Face data acquisition method, device, computer-readable medium and system

Abstract

A method, device, computer-readable medium and system for collecting human face data. The method comprises the following steps: obtaining a video file containing human face data; The data corresponding to different faces in the video file is separated by using a face tracking algorithm to generate data corresponding to different face identifiers. With the above scheme, large-scale human face datacan be collected at a lower cost.

Description

technical field [0001] Embodiments of the present invention relate to the field of face collection, and in particular, to a face data collection method, device, computer-readable medium, and system. Background technique [0002] Because face recognition technology has the advantages of convenience and safety, it is more and more favored by people, and there are more and more researches on face recognition technology. [0003] In practical research, face recognition technology based on deep learning methods requires a large amount of training data. There are currently two methods for obtaining face training data: [0004] 1. Download face data through the Internet, for example, first determine keywords such as name information, and then use a search engine to grab corresponding pictures to obtain face data. The disadvantage of this solution is that the obtained face data may contain data of different ages and different outfits, which requires a lot of labor costs for post-o...

AI Technical Summary

Problems solved by technology

The disadvantage of this solution is that the obtained face data may contain data of different ages and different outfits, which requires a lot of labor costs for post-organization, especially the beauty stars are very similar, and it is difficult to distinguish

[0005] 2. Set up face collection facilities to collect real face data. The disadvantage of this solution is that if the scale is small, the number of faces collected is small, which cannot meet the needs of deep learning algorithms. If the scale is large, the cost is too high. high

Images