/Social-Distance-Detector-

The main idea of this project is to detect the social distance between peoples at open place inorder to avoid COVID19 spread. This is done by using Python, OpenCV and yolov3 object detection mechanism.

Primary LanguagePython

Social Distance Violation Detection (SoDV) using pre-trained YOLOv3 model SoDV

COVID-19 social distancing detector using OpenCV, Deep Learning, and Computer Vision. OpenCV and deep learning can be used to implement a social distancing detector. Our configuration file used to keep our implementation neat and tidy. Our detect_people utility function, which detects people in video streams using the YOLO-v3 object detection. Our Python driver script, which glues all the pieces together into a full-fledged OpenCV social distancing detector. Apply object detection to detect all people (and only people) in a video stream (see this tutorial on building an OpenCV people counter). Compute the pairwise distances between all detected people. Based on these distances, check to see if any two people are less than N pixels apart.


Detect the social distance violations in real time using OpenCV and deep learning

"Social distancing is the only way to counter Covid-19"

Requirements

  • Python
  • Numpy
  • OpenCV
  • Scipy
  • Object Detection models (Mobilenet or Googlenet or YOLO)

Steps

  1. Input the video stream using OpenCV
  2. Detect the people in the video stream using any of the object detection models
  3. Find the centroids of the identified people and calculate the distance between the centroids
  4. Check if the distance calculated above is lesser than the minimum value (in pixels) and consider it as violation
  5. Draw the bounding boxes in red color around the person if violating, green color box if the person is following the norms

🎬 Output example:

outputimage

Frame 10 to 500:

outputimage outputimage outputimage
FRAME 10 FRAME 100 FRAME 200
outputimage outputimage outputimage
FRAME 300 FRAME 400 FRAME 500

🎯 Accuracy for person detection:

Dataset TP TN FP FN %
Oxford Town Centre 29 0 0 11 72.5

🎯 Accuracy for social distance violation detection:

Dataset TP TN FP FN %
Oxford Town Centre NA NA NA NA NA

⛔ Social distance violation detection using distance formula Drawback:

  • No camera calibration for intrinsic parameter

⏳ Future upgrades:

Features Status
Camera calibration for intrinsic parameter (distance) STATUS
Add facemask detection STATUS
upgrade to YOLO v4 STATUS

Kindly check out below links:

👀 YOLO Pre-Trained Model
YOLO Darknet