Smart City

Project Title

Smart City

Project Area of Specialization Artificial IntelligenceProject Summary

In the name of Allah, the most Gracious and the Most Merciful.

Peace and blessing of Allah be upon Prophet Muhammad ?

Cameras are everywhere around us which captures the whole surroundings 24/7. They also record abnormal activities along with normal ones, but due to huge size of data captured and human error unfortunately we are unable to detect all abnormal activities happening around us. Using computer vision and other allied technologies prove extremely advantageous for these problems. Authorities can efficiently observe and control the city using these technologies.

Computer vision plays a significant role in managing security cameras as they serve as the ‘eyes’ of the city. With the help of these sate of the art technologies we are developing a prototype which can automatically and with zero human interaction detect and alert the authorities about the abnormal activity which is being capture through cctv cameras. These activities includes person with a gun, climbing wall, throwing garbage, mud on road and stealing battery from a car.

Project Objectives

Main objectives of our project are, with zreo human interaction detect abnormal activities in captured footages, also alert the user about abnormal activity being captured so that respective actions sholud be taken. Below is one example our project detects abnormal activity in an image:

                              Before                                                                                                                After

                                     

These images are for demonstration purpose, in our project we will perform detection on real time videos.

Project Implementation Method

We are implementing 'You Only Look Once' (YOLO) - this object detection algorithm is currently the state of the art, outperforming R-CNN and its variants. YOLO takes completely different approach. It actually looks at the image just once (hence its name: You Only Look Once) but in a clever way. It outperformed R-CNN and all its variants. It divides up the image into a grid of 13x13 cell:

Each cell is responsible for predicting 5 bounding boxes. Bounding box describes the rectangle that encloses some object. YOLO also outputs the confidence score for that bounding box that how certain it is that box actually encloses an object. Score doen tell us that what kind kind of object is present in that box. The predicting bounding box may look like following, the higher the confidence score the fatter the box is drawn:

Here each cell predict two things whether or not the bounding exists and the class of that bounding box. This works like a classifier, it gives us the probability distribution over all the possible classes that the CNN has been train on. YOLO is trained on PASCAL VOC dataset, which can detect 20 different classes some of them are:

• Bicycle
• Boat
• Car
• Cat
• Dog
• Person

At last the confidence score and class prediction are combined into one final score that tell us probability that this bounding box contains a specific type of object.

The final prediction is then:

Benefits of the Project

Now a days we are surrounded with cctv cameras which runs 24/7 and also captues tens of thousands of illegal activities but unfortunately due to human error and less processinfg power we are unable to detects them. 

To solve this problem we are trying to develop a system which detect and altert the user when ever it detects abnormal activity in a video. Hope this effort of ours will play a significant role in controlling illegal activites in near future Inshallah.

Below are few examples of images in which our project detects abnormal activities:

1. Climbing on the wall

 

2. Gun(s)

  

3. Throwing Garbage

  

4. Stealing car battery

  

Technical Details of Final Deliverable

We will develop a dektop application which will directly connected with the camera and perform real time detection on the video feed coming from the camera. When ever it detects any type of abnormal activity it will generate an alarm to alert the user.

Final Deliverable of the Project Software SystemCore Industry SecurityOther IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Peace and Justice Strong InstitutionsRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	60900
MSI Geforce GTX 1080 Gaming 8GB Video Graphics Card	Equipment	1	60900	60900