lacunadream / trainlicious

Crowdedness monitoring solution using face and image recognition

Trainlicious

Background

One of the problems that many train operators have been facing is the lack of real-time passenger volume data. Taking into consideration infrastructure and cost constraints, we decided to build a solution that leveraged upon CCTV image feeds to calculate in real-time the level of crowdedness in carriages.

This repo contains the backend - reads from a postgresql database and exposes a REST API.

Issues

1. Every API call results in a database connection + call. I am not sure if the the connection needs to be recreated on every API call. Is there a way to keep the connection alive/pool properly?

Footnote

Trainlicious emerged as the champion of HackTrain 2.0

Also came in second in the post-hackathon pitching round for £25k worth of funding and a place on the HackTrain Accelerator program