SA11

SA11 (oculus)- "machine or man what's the difference?"

Comment 3

Inspiration

As we all know how pervious year was challenging for each of us especially for students in their educational aspect, sitting at home doing nothing sounds great right, well that is subjective. Sitting idle can be dangerous or productive at the same time. You tend to notice a lot of things differently or perhaps attentively, whether you look at how your dog eating his food or simply seeing a watchman dozing off. That is when you realise how important it is to push yourself and think out of the box. At the very initial stage, we started googling things and had a look at what people are focusing on nowadays. There were some interesting things which we came across, one of them is Image Classification and it got interesting. Nothing we do is ever going unnoticed. It's on CCTV cameras, it's on iPhones, it's everywhere.

What it does

oculus is an AI surveillance model which works as simply as a security guard but is as robust as a CCTV camera with various advance applications proving alarm system for burglary, fire and other panic events. This prototype model converts camera inputs into smart AI inputs. This AI mechanism mimics humanity and provides an idea about how Machine Learning algorithms can be used for image classification and other activities of an alarm system. For stating an example we can familiarize ourselves with the situation when we visit a website or simply perform any login activities, the website asks the user to confirm if he/she is a robot or not and then it tells them to select specific pictures. This model would provide a very easy-to-understand interface so that any security guard, watchman or any person who is in charge can understand it's operation.

How we built it

The programming language we chose is Python, we are using different python packages like OpenCV, Tenserflow, Keras, pandas etc. for using their pre-defined features, and we build a model that can be used. It includes ML algorithms like regression, image classification for motion tracking, all the dataset are available as google API to distinguish between the objects. For creating an energy source we have implemented the use of solar energy typically using Bi-Facial solar cells to increase productivity and consumption.

Challenges we ran into

The measure of intelligence is the ability to change. - Albert Einstein. You could claim that moving from pixelated perception, where the robot looks at sensor data, to understanding and predicting the environment is a Holy Grail of AI. Providing security is not a product but a process, saying is easing but when you do it yourself, you know how even a small error can get you in trouble and might get you in the situation when the whole prediction of the model fails. The real camera is eyes but being practical we believe CCTV camera. While building the model for the classification you have to train the dataset and test it so that it doesn't give false positives or false negatives. These two were the major challenges for us, as at last, it's the accuracy and the correctness of the model which matters the most. Dataset is huge in size so for training a model, it takes hours and when you close your computer then again you have to train it.

Accomplishments that we're proud of

Anything an individual does is for his/her gain and if it can help the community it's even great. Here as students we still are in the learning phase and we have to learn more ahead in our life so we are proud of the opportunities that all the people gave us in getting ahead in our life. Also, as a closing statement, we would like to add that CCTV will make a real difference but the human touch remains essential in cutting crime.

What's next for oculus

We are looking forward to building an easy and simple UI interface for the model so that any of our guards can understand the working and the operation because who needs a superhero when you are a CCTV operator. We are hoping ahead of that our project can be chosen for the next phase.

Built With

Share this project:

Updates