Essay

Attributing pipeline error when a detector sends the wrong crop to a species classifier

Question: Explain how to assign error in a two-stage wildlife pipeline when the object detector crops a patch of fallen leaves instead of the animal, and the species classifier outputs y=0. Describe how comparing the classifier’s output with a human judgment helps decide whether the classifier is responsible.

Sample answer: If the object detector makes a bad crop and passes a patch of fallen leaves to the species classifier, the classifier receives an input that does not contain the target animal. If it then outputs y=0, meaning no target species is present, that result is appropriate for the input it received. A human shown the same crop of leaves would also say y=0. Because the classifier is behaving correctly on the data it was given, it should not be blamed for the pipeline mistake. The error comes from the detector, which selected the wrong region in the first place.

Key points:

  • The detector passed the wrong crop, which contained leaves rather than the animal.
  • The species classifier’s output y=0 is correct for that input.
  • A human given the same crop would likely make the same negative judgment.
  • The classifier is not at fault because it performed properly on its input.
  • The pipeline error should be assigned to the detector.

Rubric: The response must explain that: 1) the classifier correctly predicted y=0 for the leaf crop; 2) a human presented with the same crop would also predict y=0, showing the classifier is blameless; 3) the detector is responsible for sending the incorrect crop.

0

1

Updated 2026-08-12

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Machine Learning Yearning @ DeepLearning.AI

Related