Torch Basics

Objectives

• Goes through the basics of pytorch in relation to computer vision
• Takes a look at components such as embeddings
• Practice on common routines

Packages to Install

pip install torch torchvision opencv-python streamlit

Computer Vision Examples

Location: computer-vision/

Basic Classification (Cats vs Dogs)

Dataset: computer-vision/datasets/cat_dog_small/ with cats/ and dogs/ subfolders containing .jpg files.

Examples:

• Build a custom dataset and inspect samples:

  python computer-vision/basic_classification/example_1_building_the_dataset.py

• Train/val split with random_split:

  python computer-vision/basic_classification/example_2_splitting_data.py

• Train a simple CNN and save a model checkpoint:

  python computer-vision/basic_classification/example_3_basic_classification.py

  This writes computer-vision/basic_classification/cat_dog_cnn.pth.

Streamlit Demo

Use the saved model from the training step to run the demo app:

streamlit run computer-vision/basic_classification/example_3_1_basic_streamlit_app.py

Upload cat_dog_cnn.pth and a cat/dog image to get a prediction.