Kvíz spojený s průvodcem · Střední úroveň
Computer Vision Engineer Career Quiz
Tests computer-vision task definitions, dataset and metric reasoning, transfer learning, role-specific qualifications, and deployment tradeoffs.
Otázka 1 z 8
Který výstup je specifický pro detekci objektů, nikoli pro klasifikaci obrazu?
Náhled každé otázky8 otázky
Odpovědi a vysvětlení se zobrazují, jak budete v kvízu.
- Který výstup je specifický pro detekci objektů, nikoli pro klasifikaci obrazu?
- 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
- Co by měl inženýr zkontrolovat po nečekaně vysokém validačním skóre?
- 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
- Kdy může být pro klasifikátor vidění důležitější recall než precision?
- 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
- Co demonstruje citovaný PyTorch Mask R‑CNN tutoriál?
- 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
- Co může inženýr počítačového vidění udělat, když je inference na cílovém zařízení příliš pomalá?
- 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
- 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
- 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
- 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