August was a very productive month for me.
Recently, I'm struggling with some technological stuff. The reason for this is because I don't know what I should learn, what I need for get a job, or what I want to keep developing.
Anyway, I want to share my knowledge and lesson in August.
I hope you'll enjoy my first entry on dev.to!
At first, I've tried to get started Kaggle. It was my first touch of machine learning.
Kaggle took the challenge called 30 Days of ML in August. When I saw it firstly, I was reluctant because I thought I can't do such ML things. However, I have an app idea using machine learning and a lot of free time. I felt it's the best time to learn ML.
The beginning started from Python grammar to actual technic like Model validation, a better way test and so on. There are a lot of terminology I have to memorize, but I can't yet.
I realized that this might not for me as a result of my through experience again. I don't like mastering machine learning, but want to use it to build an app. I knew it from the beginning but I could truly understand from the bottom of my heart and learn how to build machine learning model.
Secondly, I tried classification of the images. As I wrote the above section, I thought I don't need to build ML model by myself. Apple published a desktop app for machine learning a few years ago, It's called CreateML. Using CreateML gives you the opportunity to train your own ML model.
Pizza Classifier is a sample implementation of pizza classification app with CreatML and SwiftUI. At this time, I collected 7 types pizza images and then train and test it. Accuracy is not enough to use this app on a daily basis though, I'm satisfied with what I've learned about the way to create an ML model.
What I learned at Kaggle helped me to build ML model because I knew the reason why and how I should divide into training models and test models, at this time it's images, and how to analyze the predictions and outputs. I felt kind of connecting the dots.
GitHub Repository: https://github.com/mtfum/PizzaClassifier
Object Tracking & AR
Third project was AR app with ARKit and Vision.
Firstly, I wanted to create an app to help people who want to learn new language with showing translated words in AR world.
This basic idea depends on this GitHub Repo(CoreML-in-ARKit). It works yet, so I didn't need to build from scratch. What a luck!
The keys of this idea are an accuracy and quantity of models though, I tried to use multiplex models. However, There is no models to work perfectly.
Other than that, I know there're possibilities to build up with MLKit or some frameworks, honestly, I was getting bored so that deprioritized.
There seems to be a long path to reach out to my ideal app.
Twitter Spaces Search
This is the last project in August.
I built the basic app with taking a distance from ML.
I realized nobody tried to use Twitter API v2 for searching Twitter Spaces. Thus, I created the app which allows us to look for Spaces with a specific word.
It contains some bugs so that I wasn't enabled to finalize it in August. But, actually, I published my repository and keep to build it.
GitHub Repository: https://github.com/mtfum/TwitterSpacesSearch/
My writing skill in English is very slow like snails so the release of this article delayed by a few days. As above ideas I wrote, it's hard to make ideas into a practical and useful app. I have a lot of ideas but I feel I don't have enough skill to build up. Moreover, it would be meaningless unless nobody knows my app and progress. It's the reason I'm writing this article. From now, I want to share my thought and idea at least once a month. So then follow me dev.to and Twitter if you're interested in.
Thank you for you reading! See ya!
Top comments (0)