Facial Recognition- The Road Ahead

The argument being made is that Facial Recognition technology has a racial bias. In profiling & identifying individual targets, technology provides very accurate results in fair skin individuals (male/ female), but in the case of Black (mainly female), the effectiveness is very low and hence the issue of technology bias and related challenges.

1 article Following

In the last few weeks, 3 Global Tech companies (IBM, Microsoft, Amazon) have announced that they would not support Police dept with new Facial Recognition technology. In fact, at IBM the announcement was made by none other than the CEO, highlighting that technology may possibly get compromised in racial profiling & human rights abuse.

Recently, US has introduced a bill in Senate- Facial Recognition & Biometric Technology Moratorium Act to ban the use of Facial recognition Technology by Law enforcement agencies.           

The argument being made is that FR technology has racial bias. In profiling & identifying individual targets, technology provides very accurate results in fair skin individual (male/ female), but in case of Black (mainly female) the effectiveness is very low and hence the issue of technology bias and related challenges. This possibly could be due to low availability of data set and training module which could get addressed once the technology gets exposed to larger amount of data and with right algorithms.

FR has three major components, Computer Vision, AI based Algorithms and Database for training & validating.  The image gets captured by camera and with the help of algorithms is get mapped over the data repository to identify the target. It primarily works on Tagged feature set which builds correlation and data accumulation & aggregation. Parameters like distance between eyes, nose, shape of face, altogether 100 + data points on which data matrix gets defined.

Facial recognition/FR is very much part of our daily life, it’s a common unlocking feature for our mobile handsets, we see it at Airport for immigration and also for access control at high security establishments. Although there are other biometric scanning technologies like Iris scanner, Finger print scanner, but these technologies work more for individual identification. For large & wide spectrum of people identification FR is widely used.

Within Law enforcement agencies use of FR is very popular primarily for Mass surveillance, crowd management, people counter, Metadata for face analytics, and also for city wise dashboard.

According to Grand View Research, a Global Market research firm, Global FR market size is of US $ 3.4 B, and expected to expand at a CAGR of 14.5% in next 7 years. On the basis of technology, the facial recognition market is segmented into 2D, 3D, and facial analytics.

 Countries like China, Israel, UK, Germany are widely using FR for mass surveillance,  in China, Skynet, Government surveillance program has deployed more than 20 million Surveillance cameras embedded with FR & other video analytics applications in public areas across major cities.

There are also Security & Privacy concerns about the data collected & storage. Incidences of Data breach, cloning & tempering are well documented. There are few regulations configured recently to address these challenges. GDPR (General Data Protection Regulation) is one such regulation enacted in European Union to address issues of Privacy concern.

In additional to the Global Tech companies, lots of Video Camera OEMs offer FR analytics as part of bundled solution. OEMs from China, Taiwan & Japan command significant share of the global market. In addition, there are quite a few specialized Video analytics companies which offers FR as a stand-alone solution.

In India, share of Technology is increasing in Effective policing. Police Modernization program which was first initiated by launching Crime & Criminal Tracking Network system (CCTNS). This got further enhanced by Digital India Initiatives like Safe & Smart City programs.  As per Bureau of Police research and development- Govt of India, as on 1st Jan 2017, police to population ratio is 193 against United Nations recommended ratio of 220 (number of police personnel for 100,000 population). This ratio has substantially increased from 142 in year 2006 to 193 by early 2017. Technology could further bridge this gap and can help police in maintaining law and order, crime prevention and mitigation.

The recent Safe cities programs in Mega cities like Mumbai, Delhi, Hyderabad have components of Video analytics and facial recognition. FR Analytics & Forensics is an important element of solution deployment in these cities. This has yielded quite a good result and helped the cities in resolving major crime incidents in quick time.  Police leadership has embraced this technology which is helping them in right allocation & deployment of manpower as well as in effective monitoring.

There are quite a few Indian companies providing Standalone High-end FR solution which are locally developed. The Analytics solution deployed by these companies in Indian cities is quite robust and providing good result there by getting good feedback from the market. These companies are training their module over the local data set resulting in better effectiveness & accuracy in facial recognition. This ecosystem would be important to address the concern of local data storage. Although, in major city surveillance programs data is stored on premise but this might change in future as cloud-based solution are becoming more secure and economical and it offers more flexibility in terms of scale. So these companies would have definite advantage as compared to their global peers in terms of data storage and developing India specific unique use cases. Companies from Bengaluru, Gurugram are doing  very good job in developing & implementing state of the art technology solution.

 In conclusion, Video Analytics & FR market is poised for strong growth, it offers huge opportunity for Indian companies to build AI specific framework and develop local use cases. The benefit and effectiveness of this technology would overcome the current challenge of AI bias. The diverse & broad spectrum of solution providers would also help in negating this challenge.

In addition to Mass Surveillance application, companies working in these domains could use similar use cases for market segments like Retail and health.

Lastly, there is  urgent need for developing regulatory framework and policies for effective and diligent use of FR Technology specially in the growing markets like India. These policies surely would evolve over a period of time but the present necessity is to acknowledge and initiate immediate action in this direction. 

Disclaimer: The views expressed in the article above are those of the authors' and do not necessarily represent or reflect the views of this publishing house