Approaches to Data Labelling CT-AI Practice Questions (Page 7)

Questions 31–35

1. 31foundation · medium

   Which of the following is a disadvantage of manual labelling?

   • AIt is not suitable for tasks that require human judgment
   • BIt requires no guidelines or training for annotators
   • CIt is prone to human error and can be inconsistent between annotators
   • DIt cannot be used for image data
   Reveal Solution Discussion Select an answer first
2. 32foundation · medium

   Which of the following is a common challenge when using crowdsourcing for labelling?

   • AEnsuring consistent quality across many workers with varying skill levels
   • BThe labelling process is too slow for any practical use
   • CIt requires all workers to be in the same physical location
   • DIt is impossible to get multiple labels for the same item
   Reveal Solution Discussion Select an answer first
4. 33foundation · medium

   What is a common trade-off of automated labelling compared to manual labelling?

   • AAutomated labelling is only applicable to text data
   • BAutomated labelling is faster and cheaper but may produce lower-quality labels on complex data
   • CAutomated labelling is always more accurate than manual labelling
   • DAutomated labelling requires more human effort per item than manual labelling
   Reveal Solution Discussion Select an answer first
5. 34expert · hard

   A team is labelling video data for a model that detects actions in surveillance footage. They need to label the start and end times of each action, as well as the type of action. What is the most appropriate labelling approach, and what is a key challenge?

   • ATemporal segmentation labelling; the challenge is precisely identifying the start and end frames of each action.
   • BImage classification labelling; the challenge is classifying each frame independently.
   • CObject detection labelling; the challenge is drawing bounding boxes around the actors.
   • DSemantic segmentation labelling; the challenge is labelling every pixel in each frame.
   Reveal Solution Discussion Select an answer first
6. 35expert · hard

   A team is labelling medical images for a model that detects tumors. They have two expert radiologists and a group of trained technicians. The radiologists label a small subset, and the technicians label the rest. To ensure the technicians' labels are consistent with the radiologists' labels, what is the best approach?

   • AUse a majority vote among the technicians and radiologists for each image.
   • BHave the radiologists label a gold-standard set, then compare the technicians' labels to this set and provide feedback on disagreements.
   • CHave the technicians label all images, and then have the radiologists review all of them.
   • DUse a pre-trained model to label all images and have the radiologists review only the low-confidence predictions.
   Reveal Solution Discussion Select an answer first