Download dataset from: https://www.kaggle.com/datasets/karakaggle/kaggle-cat-vs-dog-dataset
Make sure to use only the folder "PetImages".

I've built a CNN model that is able to distinguish between cats and dogs with an 81% accuracy. Furthermore, with 10 epochs I managed to reduce the loss and validation loss significantly while preventing overfitting. Now, the CNN model can successfully predict a dog and a cat from the knowledge it has. Our dataset consists of approximately 25k images equally split among the 2 classes.

Imported modules:

Firstly, I converted our dataset into a DataFrame using pandas, with 2 columns: Images and Labels. I had to store the images as their relative paths and gave them their respective Label (0=cat, 1=dog). The next step was to split the dataset into training and testing with an 80:20 split. After splitting our data, I performed augmentation and pre-processing using ImageDataGenerator from the TensorFlow library.

Time to build our model! The model consists of 3 2D Convolution layers, 3 2D Max Pooling layers, Flatten Layer, and Dense layers. Moving forward, we define our optimizer as 'adam', loss, and metrics. This way we bring our model to completion.

The following graphs show the accuracy and loss throughout the 10 epochs.

Given such great results it was time to test our model, I chose a random image from our dataset and let the model predict. Below is an example!