Train and Evaluate a TensorFlow Classifier Model

I want you to train a Model that can classify Vehicle Types (Their Make and Model: for example = Kia Sportage)
You Can use Any one of these 2 DataSets (they are Specifically Made for Vehicle Type Recognition)
As all the images are Correctly Labeled and organized.

1)
https://github.com/faezetta/VMMRdb
2)
https://ai.stanford.edu/~jkrause/cars/car_dataset.html

Along with TensorFlow You can use any Classification Model You want (MobileNets Preferably), you can also use transfer learning no need to start from scratch.

Here is what i need.
Predection Accuracy of atleast 80%

And From tensorboard i want the following evaluation graphs
1) Training and validation accuracy after enough epochs
2) Training and validation loss after enough epochs
3) Learning Rate

Now after the model trained and ready.
I need these evaluation metrices:
1) Classification Accuracy (Test Accuracy)
2) Confusion Matrix
3) Precision and Recall Graph
4) F1 Score
5) ROC Curve and AUC

Here are two github repositories that might be helpful:
This one is very close if you want u can use it as a starting point train it and provide the evaluation metrices.
1)
https://github.com/foamliu/Car-Recognition

Or
This one contains very accurate classifier but I'm not sure if it's possible to evaluate it.
The mobileNets Classifier weights are frozen as .pb file only no Checkpoint or other files are available.
2) https://drive.google.com/file/d/13imHSThrOsrWQ7iasz-bxtPDY6dtmHK_/view