Mobile ML App development

AI Discussions
1

Hi, I’m making a mobile ML app for a project. Could I have some suggestions on how to deploy this? I’ve heard of Tensorflow Lite and Core ML, as well as Django and Django REST Framework, but I don’t quite understand how all of these work. I would greatly appreciate if someone gave me a few brief explanations.

Thanks!

2

I worked on a mobile project for a company before. We were trying to build an app to help brain trauma patients. Thus, I added an NLP Deep Learning Transformer Model. Because it can do language tasks for the patients, it can answer questions for the patients.
I ran it on my Android app, and it was fine. You need to be careful about the memory usage and the size if you really want to do offline inference as I did. Yes. that model was offline. Besides quotation, there are also other solutions that can optimize it for memory usage for Android.

I did a report for the Deep Learning model that I added for my company during my company report.
You can see it here:

Edit: I didn’t use Tensorflow. I used Pytorch for deploying it on mobile. You can also deploy Pytorch model in javascript for the web application.
You can see more examples of the Deep Learning models deployed on mobile on my youtube channel.

1 Like
3

Mobile ML App Development is an exciting and rapidly growing area that combines mobile app development with machine learning to create intelligent applications that can learn from data, make predictions, and adapt to user behavior.

When getting started with Mobile ML App Development, here are a few key things to consider:

  1. Choose the Right Platform:
    Depending on your target audience, you might go for Android (using Java/Kotlin), iOS (Swift), or cross-platform frameworks like Flutter or React Native.

  2. Select an ML Framework:
    There are several mobile-friendly ML frameworks to choose from:

    • TensorFlow Lite: Ideal for Android and iOS. It’s optimized for on-device ML.

    • Core ML: Best for iOS apps, integrates well with Swift.

    • ML Kit by Google: Offers pre-trained models and easy APIs for both Android and iOS.

  3. On-device vs. Cloud-based ML:
    On-device ML ensures low latency and better privacy, while cloud-based ML allows for more powerful processing. Many modern apps combine both, depending on use-case.

  4. Model Optimization:
    Mobile devices have limited resources, so it’s important to optimize ML models using quantization, pruning, and converting models to formats like .tflite.

  5. Common Use Cases:

    • Image recognition (e.g., object detection in photos)

    • Speech recognition or real-time translation

    • Personalized recommendations

    • Activity recognition or health monitoring

  6. Testing & Deployment:
    Always test your app on multiple devices to ensure performance and accuracy. Tools like Firebase Test Lab can help automate this.

If you’re new to Mobile ML App Development, I recommend starting with pre-trained models and using tools like TensorFlow Lite Model Maker to train custom models with less complexity.

Let me know if you’re looking for specific tutorials or code samples — happy to help!