Image classification is the task of assigning an input image one label from a fixed set of categories. This is an Image Classifier that can be used to classify images based on whether they contain a certain object or not. Image Classification is one of the core problems in Computer Vision that, despite its simplicity, has a large variety of practical applications.
I've implemented three different versions of Image Classifiers using common Deep Learning Libraries (namely TensorFlow, Keras & Numpy) for the same task in order to gain a deeper understanding of the problem and the backend working of the frameworks and their performance.
For the purpose of this project, I'll be using the Cat Image Dataset from Kaggle that includes over 9,000 images of cats which is sufficient for a binary classification task. Download the images into datasets/images/ dir and create a dir models/ for saving and checkpointing models during training.
You'll need to preprocess the images using:
python prepro.py --dataset-path datasets/images
Now you can run the train script using:
python tf_clf.py --num-epochs 10 --batch-size 64 -lr 0.001 --tensorboard-vis
Passing the --tensorboard-vis flag enables you to visualize the training loss and accuracies in your browser using the following command:
tensorboard --log-dir=./logs
- Python
- TensorFlow
- Keras
- NumPy
- h5py
- tqdm