← All workshopsRead Histogram of Oriented Gradients

Classical Machine Learning

Count edges instead of brightness

Double every direction magnitude

The edge-direction histogram has already been measured from an outline. Botie wants its directional pattern to survive a uniform contrast increase.

BOTIE

The histogram changes from [3, 4] to [6, 8]. What should happen after this length normalisation?
The inputs we are comparing

Gradient direction magnitudes

Example 1

  1. 03
  2. 14

Example 2

  1. 06
  2. 18
Result for example 1
  1. 00.6
  2. 10.8
Choose an explanation to test
Four direction bins with two empty bins

This check uses the pieces from “Remove a shared magnitude scale from edge directions”.

BOTIE

Predict the normalised direction histogram. Two bins have no gradient magnitude, but the vector as a whole has nonzero length.
Gradient direction magnitudes
  1. 00
  2. 10
  3. 25
  4. 312

Use a number, text, or an array matching the requested result. For a matrix, put each row in brackets. Saving your prediction runs your current construction on these inputs and checks the earlier examples too.

Build a construction in the workspace above first. This case will test those pieces.

Which change to a histogram does this normalisation cancel, and why would a different directional pattern still change the result?

Normalisation creates a specific invariance. State which changes it handles instead of assuming general lighting robustness.

This is a small construction problem. The lesson explains the assumptions and the parts of the full method that this example leaves out.

Work through the Python challenges →
Continue the courseKeypoints and Descriptors →
← Template Matching