What it does
Airport Assistant helps in automating the passenger entry authentication, through document-less check-in, while also assisting the passengers in navigating through the airport. It improves the travel experience of the passengers tremendously, as it simplifies the entry process and prevents the hassle of navigating through unfamiliar settings. So, all that a passenger needs to do, on entering the airport, is connect to the airport wifi to chat with the ‘Airport Assistant’ from messenger and upload a photo of their passport to validate. The Data Page on a photo ID document is generally the first page of the document with the owner’s photo and details traditionally used to verify true identity of the owner. Conventional systems depend on humans to manually verify the Passport Data Page details. The Passport Data Page Detection machine learning model minimizes a need for human intervention and automates the identity verification process using Artificial Intelligence technology. It allows a system to automatically detect if a data page is valid for identity verification by analyzing colors and patterns of the photo ID images to determine their validity in real time. There is no additional infrastructure needed in the airports and the passengers need not download new apps. This chatbot firstly checks the validity of the passport uploaded, and only after it is detected to be valid, it proceeds to show information about the various amenities including lounges, internet availability, places to eat, shop etc.
How I built it
GTRIIP's Passport Page Detection API essentially uses machine learning and AI to verify the photograph and personal particulars on the passport page. So, I deployed it on Amazon sagemaker and validated it by performing inferences first in a notebook instance. Later, I used an AWS Lambda function to invoke the endpoint, and connected it to the frontend through Amazon API Gateway. The frontend was built on messenger platform.
What I learned
I learned to use the pre-trained ML models available in the AWS Marketplace to add intelligence to the applications developed. I also learned to use the other AWS services including lambda and API gateway, which helped in connecting the model endpoint to the user-interface. I will be using them in future projects as well, given their ease of use and deployment, especially because of the connectivity between various services and the extensive documentation available for each of them.
What's next for Airport Assistant
It can be extended to provide real-time updates about flights and personalized ads based on the data extracted. Passengers who don’t have messenger app to chat from, can be redirected to a locally hosted chatbot, on connecting to the airport's free wifi.