Image classification
CNNs and transfer learning, and applying them with visual tools.
What you will learn
- Explaining transfer learning
- Building an image classifier with two or more classes
Convolutional networks (CNN) are the foundation: convolution layers detect edges and patterns, then pooling, then fully connected layers.
Transfer learning: we take a network trained on millions of images and retrain its last layer on our classes. This is how Teachable Machine (MobileNet) and the Image Embedding widget in Orange work.
Implementation steps with the tool
- Teachable Machine: an image project with three classes.
- Orange: Import Images ← Image Embedding ← Logistic Regression ← Test and Score.
Quick review question
Why does transfer learning need fewer images?
