A recent high-profile case, where a son was reunited with his real parents 25 years after being abducted at just half a year old, has once again highlighted China’s advancements and controversy in facial recognition technology.
This remarkable reunion was made possible by the technology developed by the Chinese artificial intelligence startup Beijing Deep Glint Technology. They have created a tool capable of generating adult facial images of lost babies, supporting what once seemed an impossible feat.
The “Cross-Age Kinship Facial Comparison Algorithm” was independently developed by Beijing Deep Glint. Based on genetic relationships and a relatively high similarity in facial features among relatives, the algorithm filters out suspects with higher relevance, assigns scores, and ranks them, significantly improving the efficiency of finding relatives.
In this case, the company used photos of the parents and their elder son for facial comparison. The result was used to find matches that eventually led to the lost son. According to the company, this is the fourth child it had helped the police find in the past six months.
Child trafficking has been a longstanding issue in China. There are no official statistics on the number of children abducted each year, but this month, officials from the Ministry of Public Security stated that they have found 2,609 missing or trafficked children this year. Various reports estimate that the number of children trafficked in China each year could be as high as 70,000.
The success rate of cross-age kinship facial comparison algorithms has significantly improved, allowing law enforcement to quickly identify criminal suspects and enhance the efficiency of investigations in China.
However, as a new technology for recording and storing biometric data, facial recognition technology also raises significant security and ethical concerns.
The success rate of cross-age kinship facial comparison algorithms requires a large amount of data for the machine to learn the patterns of human changes extensively. Users uploading a large amount of biometrics data could face the risk of data leakage and privacy concerns.
Additionally, because most databases of family-finding apps are relatively small, the probability of actually finding a lost family member is quite low. Moreover, there is no transparency on how the police uses such data in China.
Facial recognition technology being used to find lost relatives is nothing new in China. In 2016, Baidu launched an AI platform for finding people based on facial recognition technology. Utilizing deep learning for facial feature extraction and metric learning methods, the platform trains models on large-scale facial data and performs real-time comparisons with photos in the missing children database.
In 2018, a unit of ByteDance launched the “Face Recognition Search” feature. By uploading photos of the missing person, it can directly compare with the missing person database in the “Douyin Search for People” backend. The “Douyin Search for People” has posted over 180,000 missing person notices, connected with more than 26,000 volunteer searchers nationwide, and as of September this year, has helped find over 21,000 people.
This emotionally charged issue has even inspired a film titled ‘Dearest,’ based on a real-life story. The movie, adapted from the harrowing 24-year-long journey of a person searching for his lost son, features Hong Kong actor Andy Lau in a starring role.
