Learning Machines
Tool index
Sessions
Tool notes
Tool 04 · Session 2 · Image
What Does the Machine See?
Feature Extraction & Pixel Resolution — Session 2
Interactive feature extraction workspace
Original
Pixel-reduced representation
Representation
Pixels
Edges
Features
← Fewer pixels (compressed)
More pixels (full resolution) →
Grid: — × —
— pixels
Phone camera: ~12,000,000 pixels
Pixel values — center sample
Discussion prompts
At what grid size did you first recognize the image? Which cues were decisive — color, outline, texture, or context?
Switch between Pixels, Edges, and Features on the same image. What does each reveal or hide?
If you only had the numbers in the sample panel, could you name what the image shows?
What could a machine measure in the edge or feature map that a human might ignore?