Inspiration

Covid-19 pandemic has struck the world and affected lives of many people living across the globe. As the world slowly starts the recover from the pandemic and restart all the public places , shops, malls, gyms etc. , it is a challenge to maintain various norms to prevent the spread as people group together. The challenge of opening the gyms is that we not only have to check whether social distancing is being followed but also check that the strength of people inside the gyms are according to the guidelines issued . reducing the staff inside the gyms might pose a threat to customers as they are not being monitored while exercising. I am a regular gym goer and this challenge excited me because it is more challenging and I want to return to gyms as soon as possible.

What it does

It provides a complete safety system for gyms starting from checking whether the customers have a mask at the entrance as well as inside the gym. It will also check whether the customers are following the social distancing norms inside the gyms by calculating the distance of each customer from everyone else and highlights in red if they are not. Many countries have also imposed a restriction on the number of customers inside the gym at a given time. This maintains the count of all the customers inside the gym. The number of staff members also have to be reduced so this implies they wont be able to monitor the gym activities properly so we have made a fall detection system which will help monitor the safety of the customers even when the gym trainers are outside the exercise area . This provides a complete solution to the challenge posed by Covid on opening of gyms.

How I built it

The entire code is written in python 3.8. For person counting and person detection we have used YOLOv3 which is the most accurate algorithm present for object detection. For face detection(caffe based) and mask detection(mobile net v2) we have used opencv and imutils for preprocessing . For fall detection we used yolov3 with an algorithm based on athematics foe detecting a fall (sudden increase in width of the box and decrease in height )

Challenges I ran into

This project was full of challenges and getting the model to work accurately a was very crucial and challenging . I tried various opencv based person detectors but they failed to work accurately and finally we used YOLOv3 which performed the best. Training the model for detecting masks was also challenging and the dataset wasnt easily available. Detecting more than one person was also a challenge .

Accomplishments that I'm proud of

Since covid is a very big problem which have caused impact on gyms. With the use of this solution gym oners can safely open their gym once again without worrying about social distancing and detecting accidents in the gym without presence of trainers .I am proud of creating a code which can detect a accidents with high accuracy as it is very crucial in gym .I have learned many new things during 36 hours of my MAIS hackathon journey .But providing this solution during covid pandemic for safely reopening of gym is my biggest achievement.

What I learned

I have learned many things in image processing field which include working with yolo and open cv. Making a fall detection was a bit of a challenge .But finally we were able to build a accurate fall detector which can be used in gym.

What's next for Smart gym

We have almost made a complete solution for reopening of gym and accidents detection but further a yoga pose detector for accurate pose detection and gym workouts detector for preventing any wrong exercise can be added for accurate measurement of there poses and preventing any injury to the people can be added .

Built With

Share this project:

Updates