Inspiration
Coronavirus disease (COVID-19) is an irresistible infection caused by a newfound Coronavirus. A great many people tainted with the COVID-19 infection will encounter mellow to direct respiratory ailment and recoup without requiring exceptional treatment. More seasoned individuals, and people with fundamental clinical issues like cardiovascular infection, diabetes, ongoing respiratory ailment, and malignant growth are sure to create genuine ailment. The most ideal approach to forestall and hinder transmission is to be all around educated about the COVID-19 infection, the illness it causes and the way it spreads. Shield yourself as well as other people from disease by washing your hands or utilizing a liquor-based rub regularly, not contacting your face and wearing a veil. The first three parts need to be governed by ourselves but it can either urge people or motivate them to wear masks, the proposed project implementation has attempted to make people aware that face masks are essential for their own and other’s safety.
What it does
The proposed system develops classification and predictive model that can account for accurate classification grouping and prediction of Face masks on the face of a person. The proposed system will focus on enhancing the prediction by increasing its accuracy and detection probability. This is done by using MobileNet_V2. This system also has the ability to identify the persons who are not wearing the masks and send them a mail notification.
How we built it
The data set for the face masks are loaded into the training script. The data is then preprocessed for being fed to the classifier model. For the training purpose, a Keras/TensorFlow library named MobileNet_V2 is used, this classifier remains a better version for the CNN neural networks as in this the training procedure is relatively faster with a minimal increase inaccuracy. The training procedure when completes is stored to the disk in ‘t5’ format. To monitor the training process in this model, the matplotlib library is used to plot a graph. The OpenCV module kicks in to start the video stream, next the program detects faces in the video stream, the face mask classifier is applied to the face ROI to determine “mask” or “no mask”, the results are shown in a highlighted box around the face ROI. If a mask is detected, then the program searches for nearby faces, If a mask is not present, the person identification model starts and tries to identify the person.
Challenges we ran into
We had problems integrating the two models.
Accomplishments that we're proud of
We have published a paper on this topic..... https://ieeexplore.ieee.org/document/9297399
What we learned
We learnt working as a team and many more libraries of python.
What's next for Rebirth
The future work is as follows: - Perform the classification efficiently Using multiple datasets which could attain the optimum prediction. Database creation and addition of people in that database who are frequent defaulters Improve the overall time complexity of the entire workflow. Integrate the Person identification model and face mask detection model into a single detection algorithm.
Log in or sign up for Devpost to join the conversation.