China Tops Global Rankings in Computer Vision Surveillance Research: CSET

A recent study from the Center for Security and Emerging Technology (CSET) has shed light on China's significant lead in the research of AI-related surveillance technologies. The report, titled **Trends in AI Research for the Visual Surveillance of Populations**, delves into how China's research sector is producing a disproportionate amount of work in three core areas of AI surveillance: person re-identification (REID), crowd counting, and spoofing detection. These technologies are crucial for identifying individuals, monitoring crowds, and detecting attempts to bypass identification systems, respectively.

The study, conducted by Ashwin Acharya, Max Langenkamp, and James Dunham, analyzed a vast dataset of scientific papers published between 2015 and 2019. Their findings indicate that China's researchers are not only leading in these specific surveillance technologies but are also increasingly contributing to the broader field of computer vision, outpacing Western publication rates.

China's clear lead in research initiatives into more controversial sub-sectors of computer vision research, chiefly related to surveillance. Source: https://cset.georgetown.edu/wp-content/uploads/Surveillance-in-the-CV-Literature.pdf

China's Focus on Human-Facing Computer Vision

The report highlights that a significant portion of Chinese research focuses on human-facing computer vision tasks, such as emotion recognition, face recognition, and action recognition. These technologies, while often used for benign purposes like social media photo tagging, could also be employed by governments for more repressive surveillance activities.

The authors note that while visual surveillance research accounts for less than 10% of all computer vision research during the study period, China's dominance in both computer vision and visual surveillance research is undeniable. They state, **‘Researchers with Chinese institutional affiliations were responsible for more than one third of publications in both computer vision and visual surveillance research. This makes China by far the most prolific country in both areas. Chinese researchers’ share of global visual surveillance research is growing at a similar rate to their share of computer vision research.'**

Limitations and Broader Context

The study focused solely on English language scientific papers, which the authors acknowledge limits their findings. They suggest that including non-English publications, especially from China, might reveal even more extensive research efforts. Additionally, incorporating patent data, camera deployment, and relevant government policies could further illustrate China's lead in these fields.

The authors used Natural Language Processing (NLP) techniques, specifically the SciREX model trained on data from Papers With Code, to analyze over 100 million publications across six academic datasets. The SciBERT classifier, trained on Arxiv preprints, helped identify computer vision papers within this corpus. However, the reliance on English language documents means the study likely underestimates non-English research output, particularly from China.

Key Findings and Implications

The report found face recognition to be the most recurrent task in visual surveillance research, appearing in over a thousand papers in 2019 alone. Crowd-counting and face-spoofing recognition are also rapidly growing fields. The authors express concern that even seemingly neutral computer vision technologies can contribute to repressive systems. For instance, action recognition can detect 'abnormal behavior' in public spaces, face spoofing can prevent journalists and activists from concealing their identities, and emotion recognition can be used to identify security threats in crowded areas.

From the paper, the most frequently recurring tasks individuated for the years studied. Cited source is ‘CSET merged corpus. Results generated July 22, 2021'

The authors conclude that China's share of both computer vision and visual surveillance research has increased over time, while the United States and its allies have maintained a similar level of output. However, the global share of surveillance research from other regions has either remained stable or declined, highlighting China's growing dominance in this area.

This comprehensive study underscores the importance of monitoring global trends in AI and surveillance technology research, particularly given the potential implications for privacy and civil liberties.

**First published 6th January 2022.**