In this world of Corona virus government and private organization want to ensure people are wearing mask. They need a platform which can quickly identify whether a person is wearing a mask or not.
This can be achieved by Artificial network which will recognize if user is wearing a mask or not.
The method used in it is called Object Detection.
Object detection can be done by many methods by making your model using CNN,R-CNN or by using Microsoft Azure services Microsoft Cognitive Services. I opted for Azure where I don’t have to write entire code for making my model, I just need to train my images and model is ready but before I explain the various step involved in it I would like to clear some of the terminology in Computer Vision in a very easy terms.
Table of Contents
Image Classification Vs Object detection
As a name suggest image classification is classifying the image whether it is a dog or cat. If there are multiple images in a picture we use object detection. Object detection is a computer vision which allow to recognize and locate a object in image or video.
Object Detection using Azure ML is quite easy. We do not require big hardware or software for building a model in Azure ML. Well known algorithm can be build by simple drag and drop thus minimizing the code development.
It is also important for us to understand how determine our model performance is good or not. For this as we need to understand Precision, Recall and Map.
What is Precision?
Precision is how often my model has predicted accurately. It is proportion of correct positive result out of all predictive positive results
For example I have total 15 images and among them 9 are images which model has predicted with mask. But among 9 only 8 are correct, so my precision will 8/9 i.e. 0.889 or 88.9%
What is Recall?
Proportion of actual positive cases that model has predicted correctly.
For example in actual there are 10 images with mask but my model could predict 8 among them,So my Recall is 8/10 i.e. 0.8 or 80%
Now let dive into building a model
I created a model for detecting face mask which is quite popular in current pandemic situation.
- Collect dataset that is images of people with and without mask.
- Log in Customvision.ai using your Azure Account.
- After clicking on New project, Create new project dialogue box will pop up.
- Upload the dataset and perform the annotation on the images i.e. the labelling of images whether it is with or without mask.
- After completing annotation of images click on Train button .
- Next we need to export the model so that we can evaluate on client side. There our Five option available, I chose TensorFlow.
You can see the
performance of your model by clicking on Performance tab
In the above figure we can see a good performance of the my model.
Now you can use this model on your local system.
We help Businesses to Grow more, Grow faster, Grow Bigger using Power of Digital Transformation , use of latest IT technologies, Business Analytics, AI & ML etc.
We provide consultancy, technology and outsourcing solution to enable organizations to achieve their business objectives and long-term vision through increased IT efficiency and innovative solutions.