How do I Add multipe classification model in Android application efficiently

Question

I want to add multiple photo check features in the android app. Those are kind of image classification.

Like Office empoloye

1. Happy or Sad
2. Formal dressing or Information dressing
3. Put on mask or not
4. Inappropriate gesture
5. Sleeping or working

For each of this I need to build classification model.

When I take 100 image for happy and 100 images for sad. And build tflite model using 90x90 target size the model size is around 47mb and apk size goes around 75 mb. I may need to add more images and increase the target size to improve accuracy.

Model is created using tensorflow and converted to tensorflow lite and integrated in android app

IF I include all model it will have 5 models of size 50mb each so app size can go 75x5 MB. Which is huge.

I am mainly into data science and not goood any android developement. Any suggestion to make this architecture efficient ?

What is the best way to include 5 tflite modles in app