Cheers N.M.

Inspiration

The project is Cheers N.M. for every person who have a smartphone to know the live information about a night market event. The key inspiration for this project is from a video on Bilibili which is about a vlogger’s night market experience. I found out that there was currently no queuing time displayed in the night market, though queuing to buy things in the night market was a very common thing. The man in the video had difficulty in choosing, and he hesitated to line up to buy delicious food. My team then decided to build a project to help people who always hesitates to line up waiting for some good stuffs.

What it does

Our projects are presented in the form of web pages. The users will be able to see at a glance which merchant is the most popular. For the users who want to find the line waiting information of a specific merchant, we have an alphabetic list of all the merchants on the left of the webpage. When the users click or move the mouse onto a merchant’s name, it will show the approximate waiting time and the last update time. The most important feature is that we provide a channel for uploading photos, so that users can know the approximate time they need to queue after uploading a picture of the number of people in the queue in front. In this way, it is convenient for users to make better choices about whether to continue queuing.

How we built it

We use Javascript, jQuery, D3, intro.js and AnyChart as our front-end techniques. For the back-end platform, we chose Django to take the responsibility since it’s Python-based. The key technique of transforming a photo to the number of people lining up is computer vision, which uses machine learning to achieve the goal. The neural network that trains the dataset to detect the number of people in a photo is CSRNet (Dilated Convolutional Neural Networks for Understanding the Highly Congested Scenes) from a paper written by Yuhong Li .etc. from UIUC and Beijing University of Posts and Telecommunications. We then implemented a simple algorithm to convert the number of people waiting in line to the approximate remaining time for waiting.

Challenges we ran into

The major challenge we encountered was to generate the CSRNet paper’s model to work properly and as precisely as possible for our web. Although we have not made the prediction of the number of people in a image very precisely yet, we should master the technique of using the CSR neural network over time. Plus, due to the limited time, the whole functionalities are only available on web but not on phone. We are confident to implement the unachieved functionalities on smartphones in the future.

Accomplishments that we're proud of

We used word cloud in our UI design.

What we learned

Though three of my team are CS majored students and one is ECE majored students. We still had a hard time to convert ideas into NO-BUG code through various methods. We have looked through towns of tutorials. All of us have learnt something brand new. We also learned that we never really realize how much we are capable of doing until we give it a shot.

What's next for Cheers N.M.

Next, we will continue to optimize and perfect our project. Apply all functions to smartphones. At the same time, we will also work hard to improve the accuracy of queuing people identification. If it works well, it may cooperate with the night market organizer. Stay tuned for more features!