一Motion

• 

  Prototype of smart system

Inspiration

Singaporean students tend be afraid of asking questions, despite being unsure. On top of that, tuition spending doubled from 2004 to 2014. Both trends inspired our team to come up with 一Motion, a device to capture student's level of comprehension during class. This hope to empower teachers with the ability to reallocate their attention to weaker students while focusing on the delivery of the class.

What it does

一Motion leverages on advanced technologies to capture emotional data to obtain student's level of comprehension. This data is processed and aggregated to provide immediate feedback to teachers such as slowing down or repeating a topic. After class, teachers can review the results and determine which students need more attention or remedial classes.

How we built it

There is 4 main modules we had to build. Firstly, the device to capture emotional data which is later processed. We made use of a raspberry pi with a camera module and Microsoft Emotion API, we programmed it to take photos, upload to the API and store emotional data. Secondly, the data is then uploaded into Azure SQL cloud which can be accessed later for machine learning. Thirdly, we created the training data by taking hundreds of a teammate's photo and analyse it to create our machine learning model, using Bayes Algorithm. Lastly, in order to make insights actionable, we created a dashboard for teachers integrated with our machine learning model. This dashboard provide relevant insights and recommended actions so that teachers can focus on the delivery of lesson while providing personalized attention during/after class.

Challenges we ran into

It was difficult to find a SQL cloud server to host our data such that it can be easily integrated by our application. Also, the design of our product has to be not too intrusive such that it distracts the student. One photo of emotional data alone does not provide sufficient information whether a student is confused or doing something else instead.

Accomplishments that we're proud of

We managed to successfully trained a data of 80% accuracy which is usually quite difficult to achieve with real-world data. Also, despite coming from different backgrounds, we managed to work closely as a team and meet our own goals.

What we learned

Technically skills alone, we learn how to integrate APIs, how to program with Raspberry Pi and creating an effective dashboard. Whereas for soft skills wise, we learn how to manage our time better and importance of project planning to maximise our efforts and limited time.

What's next for 一Motion

After having a proof of concept to demonstrate its capability, we hope to further develop it by improving on its existing functions and upgrade it to be just one camera to capture the whole class. This will therefore minimise the intrusive element of our product.