Case Study

Attributing a correct detection with a wrong subclass label

Case context: You are building a two-stage system that first detects fruit in orchard photos and then classifies the fruit variety. In one test image, the detector draws a tight bounding box around an apple, but the second stage labels it as a pear.

Question: Which component is responsible for the mistake, and how should the error be attributed?

Sample answer: The fruit-variety classifier is at fault. The detector succeeded because it found the fruit and produced an appropriate bounding box. Since the location step was correct, the incorrect final label must come from the classifier stage.

Key points:

  • The detector correctly located the fruit with an appropriate bounding box.
  • The classifier produced the wrong variety label.
  • The error should be assigned to the classification stage.

Rubric: The learner must identify the fruit-variety classifier as the source of the error and justify this by stating that the detector correctly produced an appropriate bounding box.

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