Pytorch image classification custom model

Job ID: 32162933

Budget: $30 – $250 USD

I am a machine learning engineer and I need to with the following:

Task: Classify pills

Challenge: All my models tend to over fit.

What needs to be done:
1. Build a model in PyTorch that can reliably classify the pills without over fitting
2. Explain why your model architecture performs well

Success criteria:
- The model needs to score %95+ on an unseen test images

What will I provide?
- The data-set of pills images and the augmentation logic I used. You are free to use any augmentation logic you want.

Limitations?
- No limitations as long as (1) you use PyTorch, (2) you can explain why your code works and (3) your model scores %95+ accuracy on unseen data that I have