Saffree - The First Sprint

In this 2 week long sprint my scrum team and I have managed to accomplish the main features of the project, including the NRIC scanner, tracking of visitor's movements (whether the visitor entered of left the venue), turning a video from an angled perspective to a birds-eye view, as well as the main software for the programs to communicate with each other through the use of sockets. Our group has also managed to complete the design of the final product, and made a prototype using cardboard boxes. Even though the system is working with the most basic features, it is still very rough around the edges, with some features still incomplete.

For this sprint, my responsibility was to make use of object detection and object tracking to detect humans going in and out of the store, detecting when someone is approaching the "Saffree Box" and scanning the visitor's NRIC. I also had to link the various programs together using sockets,

The above video is a quick glance on how the tracking of visitors will work. The program is currently running on a Raspberry Pi, hence I chose to use a lighter object detection model which uses less computational power. The program searches for human objects every few frames, and tracks the detected human object afterwards, this is to improve the efficiency, since the Raspberry Pi cannot handle object detection on each and every frame while tracking the object's position.

As for the detection of visitor's approaching the "Saffree Box", I made use of a lidar sensor to detect the distance, and when the distance reaches a certain value, the program will activate the camera, which will then detect for an NRIC barcode. The system will then record the barcode, and communicate with the server that someone has scanned their NRIC. By doing so, the visitor is allowed to pass through the entrance without triggering the alarm. However, if the tracking camera detects that the visitor entered the venue without scanning at the "Saffree Box", the alarm will be triggered and the visitor will be reminded to scan before entering.

During this sprint, I have faced various issues, and most of it is with regards to the efficiency of the program. The future goal of this project is for the program to read from a live video and make updates in real time, however, this cannot be accomplished without efficient code and a well-trained model. Even though the current program can work decently well with a pre-recorded video, it struggles with a live video feed. In the future sprints, this will be solved by training our own model using our own dataset, which will remove the redundant object detection classes the current model has. There is also the possibility of using an IP camera instead, so that the program can be run on a remote server, greatly improving the performance and accuracy of our system.

This sprint was carried out using various scrum ceremonies. Each P2 lesson, my group had a stand-up scrum meeting, where we discuss the issues each of us faced and solved any issues discovered. When this sprint was over, we reviewed the sprint, and also looked back at what went wrong, and what went well as part of a sprint retrospective. This allowed us to learn do better in future sprints.

Sprint 1 Backlog

My scrum team worked well together for this sprint, and gave each other a helping hand when issues arise, we met our sprint's target, and managed to do well in our first project pitch after the 2 day long hackathon. And in our future sprints, our current system will be upgraded and improved, while adding new features to improve the Saffree experience.