Inspiration
The main inspiration behind the solution was to deliver a simple solution, addressing the two main aspects of hygiene during the pandemic. It mainly focuses on images and videos taken from a distance (eg: CCTV videos , distant photos)
What it does
There are lot of other solutions that focus on the same but this project is mainly for videos and images taken from a distance, eg: mainly taken from a CCTV camera. The solution consists of two main parts.
- Social Distancing : In this part the criteria of maintaining 3ft distance is checked and an output is given
- Face mask/ no mask detection : In this part a simple detection model is used to check if the person is wearing mask or not.
How we built it
The solution is built using python, tensorflow and uses. - YOLO - Object detection using Faster-RCNN The solution uses open source models (YOLO and Faster-RCNN) and these models are trained using custom data to give apt results.
Challenges we ran into
The main challenge was the images taken from a distance and face not being clear. The main challenge for social distancing was people not being clear in the image.
What we learned
We learned that for any system to be built properly, proper and apt data is necessary
Applications Safe-Enviro
Safe-enviro can be used by anyone. It can be installed in your houses, companies, garages etc.
Referral Links
The open source model and code used for Faster-RCNN was taken from this Github link: https://github.com/EdjeElectronics/TensorFlow-Object-Detection-API-Tutorial-Train-Multiple-Objects-Windows-10 The darknet model from this link: https://github.com/AlexeyAB/darknet
Built With
- anaconda
- neural-network
- python
- tensorflow
- yolo
