Many newb criminals are getting arrested and they tend to make more mistakes under the influence of gangsters and they become evil to the society which in turn affects the country. In order to change that situation we took up this problem and worked upon it.
What it does
It recognizes facial and gesture recognition and if any mischievous activity is taken place he'll be caught by emotion recognition owing to the face agitation as it is caused by the new criminals and so they'll be not much of a pro in this field. It'll give credit score to each one of the prisoner and it'll encourage them to be good in order to get advantage over the other and also a quick release or something like that depending upon the law.
How we built it
We built it using Flask Framework which involved haarcascade model for facial recognition, gesture recognition for cost efficient purpose and contact less activity due to this pandemic as prisoners are also human we need to take care of their safety too and emotional recognition for face agitation. The front-end was built by HTML, CSS, JS and BOOTSTRAP. The admin and warden can look into the prisoners activity whenever needed.
Challenges we ran into
Retrieving the images. Setting up the threshold limit for each step we had to run many tests because it ran into to give less accurate results. And after many trial and error methods we successfully got the threshold time for each step as face, gesture and emotion recognition. We ran into a problem of data flushing for the next prisoner it logged in as the same prisoner as of last time. We spent more than 2 hours on finding the bug and we recovered it by defining every functionality of the project in separate functions and routing it to app.py (flask) and flushed the data successfully.
Accomplishments that I'm proud of
With this system we would be giving out good citizens who come in as bad citizens. With this Application we are able to control and regulate the prison inmates more effectively with very low cost. This Application makes wardens work more easier and it also encourage the prisoners to follow the rules and regulations of the prisons.
What we learned
Kiran Subramanian S – Learnt a bit extra in backend and ML. Naveen Kumar B – Leant a bit extra in frontend and PySQL. Nithin K M – Effective use of git and debugged the team’s mistakes. Vijayaalayan A– Haarcascade model and of course Video editing for the submission.
What's next for Tracker Mate
Violence detection : We will develop a model which detects the fighting activities and report to the warden immediately or sound an alarm.
Crowd Detection : We will develop a model which would detect specific person from a crowd if he performs any illegal activity like weapon exchange/ drug exchange.