Women's security, for some, may seem like an overused issue. So, let’s come straight to the point, every year the National Commission for Women gets thousands of complaints on crime against women. And in 2021, it was an all-time high of 31,000. It is also to be remembered that only a very small percentage of violence against women actually gets reported. With the success of movements like “Me Too,” more and more are raising their voice now. But the unfortunate truth remains - society needs to act fast. With deep tech transforming nearly every sector, can AI come to the safety of women?

Meet Souvik Ghosh, at the age of 21, he teaches sex workers’ children of the Kalighat red light area in Kolkata, West Bengal. Currently, he is a 3rd-year student pursuing BTech from Heritage Institute of Technology, Kolkata.   

Souvik is not just making a difference with education but is currently working on an AI project aimed at women security named ‘Sathi.’  

“I am not just working in the data science and computer vision domain, but I am an active social worker as well. I teach sex workers’ children on behalf of the Rotaract Club of my college. I have seen sexual abuse a bit more closely than anyone can, and even children are exploited daily,” said Souvik. 

Probable ‘AI Sathi’ for women 

Sexual harassment with women is common at busy roads, schools, colleges, railways, markets, or any other public place one can point to. Most of them either get too shocked to react or feel embarrassed to shout/protest to avoid creating a scene. These very instances made Souvik look for the solution and soon a thing came to his mind - one thing that is common at all these places is the presence of CCTV cameras. 

Explaining his project Sathi, Souvik explained: “Imagine a situation where he/she is feeling unsafe and facing some form of harassment, but unable to shout. Now, the victim needs to stare at the camera, show her fist - it will play a loud alarm (embedded with CCTV cameras) in response to the fist.” 

While making the face, if the woman shows her index finger, it will point as number ‘1’ for the system which in turn will contact the nearby police station. Also, every facial expression will be captured by the camera and will be updated in the database to be used as evidence later.  

Tech behind 

Souvik, in the past, has worked on applications involving the use of AI in healthcare, business, agriculture. The current AI project has its roots in deep learning and computer vision.  

• Sathi takes the help of MediaPipe, an open-source platform from Google for building multi-modal ML pipelines.  
• Hand movement detection is done with the help of the Python language. 
• For the face recognition part, an entirely new model is built from scratch in Yolo. This is to recognise faces even in the low light areas, which is not the case when it comes to MediaPipe, or other libraries available. 

Challenges at hand 

Souvik is confident about his project, right now a prototype, to become a full-fledged technology that can be employed alongside CCTVs through means of embedded chips, or to create a separate app for smartphones especially when it comes to rural areas where CCTVs are rare. However, Souvik seeks help from the experts in the AI domain to resolve certain challenges, some of them as rightly mentioned by him is stated below: 

• Firstly, it is required to know what distance the system can work accurately. If the system can work for 10 m, 15 m, or can it capture the data accurately over 20 m? 
• Secondly, how well is the system’s performance when it comes to a crowded place, such as a market or a local train? Currently, the system works well for up to 10 people, but how the system makes decisions in densely populated areas such as Delhi, or Kolkata for instance. 
• The third and the most important part is what if people start showing their fists just for the sake of fun. How can the system differentiate a genuine case from a prank?

The AI-based ‘Sathi’ project may just be a prototype as of now, but the project holds promise. Experts across the AI domain can come and help this young 3rd-year student to convert the well-thought-out project to be a reality for strengthening the cause of women’s security.