Útmutatóhoz kapcsolódó kvíz · Közepes Szint

Computer Vision Engineer Career Quiz

Tests computer-vision task definitions, dataset and metric reasoning, transfer learning, role-specific qualifications, and deployment tradeoffs.

Kapcsolódó útmutató utakSzámítógépes látásmérnöki karrier
kérdés 1 -ból 8

Which output is specific to object detection rather than image classification?

Tekintse meg az összes kérdést8 kérdéseket

A válaszok és magyarázatok megjelennek a kvíz kitöltésekor.

  1. Which output is specific to object detection rather than image classification?
    • Classification requires bounding boxes for every object
    • Object locations or bounding boxes paired with class labels
    • One or more labels for the whole image
    • A pixel-level class mask for each object
  2. What should an engineer check after an unexpectedly strong validation score?
    • Whether the model has enough deployment replicas
    • Whether the color profile matches the dashboard
    • Whether the optimizer uses a larger learning rate
    • Possible leakage or overlap between data splits
  3. When could recall matter more than precision for a vision classifier?
    • When missing a positive case is costlier than a false alarm
    • When false alarms are the only harmful error
    • When the dataset has no positive labels
    • When all thresholds produce the same predictions
  4. What does the cited PyTorch Mask R-CNN tutorial demonstrate?
    • Replacing all image labels with embeddings
    • Deploying an autonomous vehicle system
    • Fine-tuning a pretrained model for detection and instance segmentation
    • Training a universal vision model from scratch
  5. What can a computer-vision engineer do when inference is too slow on a target device?
    • Change the evaluation labels until runtime improves
    • Choose a different metric without measuring latency
    • Assume cloud inference is always acceptable
    • Profile the pipeline and evaluate model or deployment optimizations
  6. What qualification does the cited Google posting list for its specific role?
    • Programming experience in Python or C++
    • A required doctorate for every vision engineer
    • Experience only with JavaScript
    • A guarantee of prior autonomous-vehicle work
  7. How does image segmentation differ from object detection in the guide?
    • Detection assigns a label to every pixel
    • They are identical outputs with different names
    • Segmentation assigns labels at pixel or instance level
    • Segmentation returns only one image-level class
  8. How should candidates treat the guide’s practice prompts?
    • As replacements for reading job requirements
    • As preparation examples, not a promise about an employer’s interview
    • As leaked questions that all companies use
    • As a required coding test for every role