Build 2 models ( No. 1. transfer learning- completed but to be adjusted and No.2. CNN- to be done ) for comparison.

Work Instruction:

General instruction: You should do transfer learning first ( to mend code available on github -accuracy is 94% ) and then you should develop/train your enhanced CNN.

Then you should do a comparative analysis of both model.

Questions I will be having for you when you are catering for the below in your python codes?

-Application of pre-trained model

-Development of enhanced CNN

-Explanation on training set and testing set

-How you have trained the model and solve the issues of overfitting

-There are very little difference between different abnormalities. How you cater for that?

-What are the parameters that influence the performance

-Detailed description on performance and evaluation

-In medical field, GAN is being used instead of data augmentation -What you do in case of many unlabeled data?

Reference No.1 for Transfer Learning:

Tutorial: [login to view URL] watch from 33:15 to 43:15

Github code: [login to view URL]

Reference No.2 for Transfer Learning:

Github code: [login to view URL]

Reference No.3:

Github code: [login to view URL]

Note: You can use google colab for the training of the 2 models.

Payment clause: Payment will be done when both models have been developed and trained and full comparative analysis reports are done

Kemahiran: Python, Deep Learning, Image Processing, Machine Learning (ML)

ID Projek: #31562133