Inspiration

PROBLEM Child pornography is criminal offence and a repercussion of child trafficking. Tens of thousands of videos are uploaded on many websites throughout the world in a day. (70,000 urls hosting child pornography were hosted in 2015) Filtering videos on Porn websites manually is very slow process which leads to gateway for people to easily upload videos which are against such law. This is a global social problem which needs to be addressed using latest technologies. SOLUTION By analysing the content of video before it get published, our product checks on such restricted content and discard the upload if found. We have developed an video analysis platform, which internally works on advanced image recognition using machine learning and neural-network concepts. It recognizes if a given pornographic video has child pornography The video is then further reported to the website/manual testers for recheck. And if the child pornographic contents are found in the video it is then banned. DATA IS THE NEW OIL Keep a record of metadata of videos being uploaded ; IP address ; city; zipcode ; age ; email address of child porn uploaders Keep a record of the unerage personnel identified for further reporting to suitable authorities Data further used for maintaining a central database which could be shared with corresponding authorities on a freemium basis. THE NEED Indian Govt banned ALL porn websites in 2015 on account of child pornography getting discovered on a few websites leading to a loss of business for legitimate and legal porn websites Youtube pays 5000$ per month to unemployed new graduates for human moderation.[1] Lot of human moderation is also outsourced to Phillipines for instance costing upto 500$ per month for human moderation[1] Human moderators face psychological problems TRACTION Reached around 50 porn websites with this solution and many of them reciprocated positively as human involvement is video analysis is psychologically tedious job.

What it does

By analysing the content of video before it get published, our product checks on such restricted content and discard the upload if found. We have developed an video analysis platform, which internally works on advanced image recognition using machine learning and neural-network concepts. It recognizes if a given pornographic video has child pornography The video is then further reported to the website/manual testers for recheck. And if the child pornographic contents are found in the video it is then banned.

How I built it

With opencv,java,machinelearning,ffmpeg

Challenges I ran into

running it with limited ram of 1gb on free aws version was by far the biggest challenge. It was gracefully handled my dynamically allocating stack size to threads as per their requirement

Accomplishments that I'm proud of

Opencv wrk, java thread executor service, machine learning work

What I learned

thread executor service, java memory handling, computer vision

What's next for #Pragma

Going to make it even mor better and pitch it to potential mediums where it can be used effectively for solving this social problem

Share this project:
×

Updates