Training Custom Image Models Quiz

5 questions Pass: 70% +25 pts (50 with Pro)

Quiz covering Custom Vision Models

What you'll be asked

  1. What is the primary purpose of labeling images when training a Custom Vision model?
  2. You are training a model to detect defects in manufactured parts. You notice that the model performs well on training data but poorly on real-world test images. What is the most likely cause?
  3. When building an object detection model, why is it important to include 'negative samples' (images that do not contain the target objects) in your training set?
  4. If you have a dataset with 1,000 images of 'Cats' and only 50 images of 'Dogs', what is the most effective way to improve the model's ability to identify 'Dogs'?
  5. You are using a transfer learning approach for a Custom Vision project. Which of the following best describes the benefit of this technique?

Answers and explanations come after you submit.