Case Study

Prioritizing engineering efforts in a sequential image processing pipeline using component error counts.

Case context: An engineering team is analyzing a multi-stage pipeline consisting of a cat detector and a cat breed classifier. After inspecting 100 misclassified dev-set images, they attribute 90 of the errors to the cat detector and 10 errors to the cat breed classifier.

Question: Based on this analysis, which component should the team focus more attention on improving, and why?

Sample answer: The team should focus more attention on improving the cat detector. This decision is justified because the cat detector is responsible for the vast majority of the errors (90 out of 100 misclassifications), whereas the breed classifier only accounts for 10 errors. Targeting the cat detector provides a much greater opportunity to resolve overall pipeline failures.

Key points:

  • Focus development attention on the cat detector.
  • Identify that the cat detector causes 90 of the 100 dev-set errors.
  • Contrast this with the breed classifier, which only causes 10 errors.

Rubric: The answer must recommend focusing attention on the cat detector and justify this recommendation by noting that it is responsible for 90 of the 100 misclassified images, while the breed classifier is responsible for only 10.

0

1

Updated 2026-05-27

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