Starting Project with Pre-Trained Model

Hi Mentors,

I have finished ML Specialization and DL Specialization. I now want to implement a small project using a pre-trained model (from model hub such as Hugging Face) to increase my confidence.

Don’t know how to begin with first creating the model and then how to wrap it for an android device and how can I use a platform for free to deploy it. Please guide step by step in brief.

Hi Nitin!

First, decide on your problem statement. Collect the data you need for that or else if you don’t have a problem statement and just want to learn, try doing experiments on popular datasets like mnist datasets etc because it will be easy to handle and you will get lots of help on the internet if you get stuck. This will boost your confidence. Don’t directly jump into complex problems for which you might not get any help.

To deploy it in Android, you cannot directly use your trained model as it will be computationally heavy, you might be required to convert it into a lighter format using TensorFlow Lite or PyTorch mobile and also you would be required to know some level of Android development (Ex: Android Development using Java or Kotlin or React-Native etc). Check out the TensorFlow: Data and Deployment Specialization, it might be helpful in teaching you how to do these kinds of stuff (deployment). Deployment is somewhat a different kind of art compared to traditional model building and evaluation. It might take some time but worth it!