How does Tensorflow’s CNN compare to a pretrained CNN?

In an earlier post, I made predictions on the classification of a dataset of cheetahs and lions using a pretrained model. In that post, I used the pretrained convolutional neural network (CNN), MobileNet, to classify images images as being either a 0 for a cheetah or a 1 for a lion. When I made predictions on the test set, I achieved 86.67% accuracy rating.

The link to my previous post can be found here:- https://medium.com/mlearning-ai/how-i-used-the-mobilenet-pretrained-model-to-classify-lion-and-cheetah-images-2d97cf9f4453

In this post, I want to compare the performance of Tensorflow’s made from scratch CNN and see how it measures with MobileNet’s pretrained CNN.

In this algorithm, I removed the scaling from the preprocessing portion of the program and inserted it directly into the first layer of the model. I prepare this method of scaling the data because the researcher does not have to worry about whether the data is scaled before he inserts it into the model.

The model I created from scratch, using Tensorflow, can be seen below:-

I defined the model as being sequential. In TensorFlow, the Sequential model is a way to create a neural network by simply stacking together different layers in a sequential order. This is a very popular and…