Pulling some old story for a new blog post…
Back in March, I built an app and submitted to Google PlayStore. The app is SnapSpeak, which is an app that contains 2 main feature:
- Extract text from images
- Convert text to speech
On the other hand, the main features of the app are built using Google Vision API to extract text from images, and Amazon Polly to convert text to speech. I think these APIs are definitely good enough for simple app like mine, and it really saves a lot of time from deploying your own machine learning solution / API.
If I recalled correctly, the whole development process took about 5 days from the initial create-react-native-app to publishing in PlayStore. Not too bad! Of course, the app is very basic now and there is a lot of improvement and proper design that can be made. For example, my initial thought was to wrap these features to be an audiobook app, where you can snap photos of articles and listen to it on the go. Guess I don’t have the time for that, lol.
In the meantime, if you are interested to see the app in action, here is the video of the app walkthrough:
So other than this, what have I been up to recently? One thing, I am still trying to finish the part 2 of fast.ai course, where there are more details about the usage of PyTorch and custom neural net models implemented from research papers. The fast.ai library is good, just that it’s too abstract and I prefer to have a peek into the underlying structure of the model I am using. On the other hand, since I have trained a model for the dog breed classification in one of the fast.ai lessons, I am exploring wrapping the model with an API and expose it to be used in a Facebook Chatbot. It’s already 70% there and hopefully I can write up another blogpost about that during this month!
Till then 🙂