QHacks 1st Place Winner Q&A + Demo
Week days' dinners just got easier with CookHack using Forge
This week on Tinyverse, our team had the great pleasure of inviting the 1st place winners of QHacks for an exclusive Q&A! It was a delight speaking with Logan Hoogendijk, Hunter Hoogendijk, Kevin Yu, and, Sam Thibault who are currently first and second year students at Queen’s University! Thank you so much for making the time to chat with our team. We’re ecstatic to hear you that you all had a great experience overall using Forge!
Our 1st place winners created an amazing app to use a food model to recommend recipes for all those needing help deciding what’s for dinner! Check out the neat CookHack demo above.
What has been your experience in building ML applications?
I have built two machine learning applications, both times with a group of 5 people, including myself. The first is a computer vision model using K-Nearest Neighbours that predicts a handwritten digit's accuracy based on the MNIST dataset. The second is another computer vision model using Inceptionv3 that classifies an animal-based off on a picture.
I have developed two machine learning applications with a team of 5, including myself. The first application was a prediction software using binary classification models to analyze the labels and data of COVID-19 to predict stock market changes. The second machine learning application is an NLP model that utilizes news articles to predict stock market changes.
I did not have any previous experience in building ML applications. Exploring Forge was the first time I was introduced to EdgeML deployment.
I don't have any previous experience in building ML applications; however, I'm currently taking an ML course on Udemy and learning the fundamentals such as linear regression, and reinforced learning.
Have you deployed any ML apps to the phone?
Currently, we have not deployed any mobile applications.
Describe the idea and project you built on Forge:
The project we built on Forge is known as CookHack. CookHack is a full-stack web application that allows users to log in to a personalized account to browse a catalogue of unique recipes from our database and receive simple step-by-step instructions on cooking delicious homemade dishes. CookHack also provides the ability for users to add the ingredients that they have readily available using a picture with image detection and start cooking recipes with those associated ingredients. Lastly, CookHack encourages the idea of interconnection by allowing users to share their cooking experiences online with each other.
What was the overall experience using Studio?
The overall experience of using Studio is that it allows for seamless deployment of ML models to various applications. Also, Studio’s user-friendly UI made the process of deploying an application extremely easy.
What are some ML apps that you've used in the past that you also want to build?
We would like to build some ML apps, including developing a traffic prediction ML model that helps predict the best time to travel on the highway.
After using tools to convert scanned notes to digital text, we thought of building an application that used text recognition models such as the East Text Detector available on Forge to suggest cable layouts for custom PC builds.
Share about your knowledge around edgeML and brainstorm new ideas you have for your applications.
Our team was interested in implementing edgeML in CookHack since it greatly simplified the project stack. Deploying the recognition model locally via Forge addresses fundamental privacy concerns by mitigating the need to call external APIs.
Our application's new ideas were implementing an ML model that could recognize multiple ingredients in a photo rather than a single ingredient. Another idea was migrating our web application to a mobile application to allow users to use our application anywhere, anytime.
What were some highlights using Studio? What were some challenges?
Some highlights of using Studio was the great UI that made it extremely easy to build and deploy the model. Additionally, Studio reduced the time it takes to develop the ML model allowing for more time to create other functions during the hackathon. Some of the challenges included incorporating the deployment key into our application as we used a different bundle than the template.
In your words, what does Studio mean for developers? Can you share what you see Studio is solving?
With machine learning applications significantly gaining traction over the past few years, Studio will solve the problem of developers being unsure how to deploy models to their applications. This is done by creating a simple and effective way to implement ML models into applications with only a few lines of code. Developers will now have the option to discard the significant amount of time it takes to develop a working ML model and instead use Studio to deploy the models to their applications instantly.
Now, you can also build your next idea using edge ML using studio. Studio lets you drag and drop using out of the box models to build fully featured apps. Try it https://studio.hotg.ai or share the link with any application developer that might find it useful!