I think deep learning is accessible enough now that if you know how to program, you know how to get started using it for your own tasks. This great article shows you how you can use tflearn, a TensorFlow based Python library to create predictive models based on the CIFAR-10 dataset.

This is a simple implementation using the same neural network layout to identify labeled photos of my cat.

To run this, first create an Anaconda environment based off the environment.yml using Python 3.5. Then, create a folder images in the local directory with two subfolders cat and not_cat. Sort through your own files and copy your cat photos into cat and your non-cat photos into not_cat.

To run the training step, run:

python cnn.py

which will read all of the files and train a network based on the image features. That script will also write to a file cat-classifier.tfl which is a binary representation of the trained model that you can use in later scripts.

To use a trained model from cnn.py to classify your own images, run:

python classify.py <image_path>

where <image_path> is the path of an image you want to classify. The output from this is a JSON object with probabilities for cat and not_cat scores.