One of our team members, Bhavya had won the grand prize at an APAC-level hackathon and he got a chance to go to a startup conference in Barcelona this summer. But due to the pandemic, it was canceled. Now, Bhavya is not sure if he’ll go even if the conference is next year. We realized this would be the case for many travelers, both leisure and business. One of the major problems would be making travelers gain the confidence to travel again. We decided to do something to encourage people to travel, by assuring them of safety.
Hygiene & Health Hack the very definition of travel in ways that address the health concerns long after the COVID-19 slowdown has passed. These inventions must demonstrate how the well-being of travelers and businesses is preserved or improved during travel rather than put at risk.
Sustainability & Relief With pre-pandemic travel practices as your starting point, and new business models and social/environmental impacts as your guide, invent new responsible, socially impactful products and services that drive exceptional travel experiences and economic growth. Also, with public health and employment concerns at the forefront now, demonstrate how your invention measurably enhances economic conditions in regional and local destinations around the world.
What it does
We have created software services for hotels, airports, parks, restaurants, museums, theatres, and other enclosed private tourist spots. Our system will automatically detect whether people are following social distancing and whether they are wearing masks or not, from CCTV footage. The owners of the place will be alerted if someone is not following the rules. These places can advertise that they’re using an automated system to ensure safety, and this will attract more tourists. The other facet of our solution is an Android app for travelers/tourists. Users can pick a destination and a date of interest. We will show them the updates of that area, and give the estimated number of cases. This estimation is based on a predictive ML model. This will help users make an informed decision and they can postpone their trip well in advance, without losing out money on cancellation charges. This will also help air travel companies and hotels, who have to bear losses if a person cancels their stay.
How we built it
We have taken a sample recording of the CCTV camera footage. An image processing model detects and classifies various bounding boxes based on the distance between people in the video. Also, we have the Mask detection algorithm which was built using CNN and it checks whether people are wearing a mask or not and creates a bounding box around the face. So the viewer knows the number of people violating the norms. These models were built in Python.
The website for private owners (of the hotel, market, tourist spot) was built using React, Firebase, and Node. The mobile app for users (tourists) was built using the Android Studio and Firebase.
Challenges we ran into
Data Privacy was the one challenge we encountered during this Hackathon. However, the service which we will be providing will enable the private tourist places to analyze the data and count the number of people maintaining the social distancing norms without giving any private information of the person from the CCTV frames. In the future, we plan to add a pipeline which will blur the faces of the people present to provide an even safer and secure service.
Accomplishments that we're proud of
We are proud of the fact that our project will help many travelers, both leisure and business in the aftermath of this pandemic. We will be providing one of the major strengths to travelers that is gaining confidence to travel again. We are really happy to be part of the change that will boost & encourage people to travel safely.
We have used Machine learning to detect if people are following COVID19 norms- social distancing and wearing masks. We have used the YOLO model to detect people. Once that task is achieved we have calculated the euclidean distance between the bounding boxes(output of the ML model). We check if the distance has been above a threshold that we have defined. If so, we attribute such a group of people with a red bounding box. For mask detection, we have made a custom CNN model wherein we have annotated the mask image and trained the convolutional neural network for the mask image. The model detects the presence of masks in the frame and draws a bounding box around it.
For COVID19 trend prediction, we have used data of a place named Mumbai in India. We have used Recurrent neural networks(RNN) for such prediction because of the fact it can remember past trends and predict future trends on the basis of past knowledge.
For the website, React, Node, and Firebase were used. For mobile apps, Android Studio, Google SDK, Covid19 API, and Firebase were used.
What's next for EZTravel
The next plan would be to host our entire application on the cloud. The ML models and the backend will be deployed on the cloud. In phase 1, we would like to try out this solution locally. We will tie-up with local hotel chains and tourist spots in Mumbai and devise a basic billing plan to start earning revenue. We will also release our app for tourists on the play store. After these iterations and learning from the results, we would like to partner with more places and or a company like Trivago which can in turn sell these services to its partners.