CNN Object Detection Models 2 — Questions and Answers
Question 1: What is the primary role of the Region Proposal Network (RPN) in Faster R-CNN?
- Generate class probability scores
- Propose candidate object bounding boxes (Correct answer)
- Perform non-maximum suppression
- Extract feature maps from the backbone
Correct answer: Propose candidate object bounding boxes
The RPN slides over the feature map and proposes candidate bounding boxes (anchors) that may contain objects.
Question 2: Which loss function does YOLO use for bounding box coordinate regression?
- Cross-entropy loss
- Focal loss
- Mean squared error (MSE) (Correct answer)
- IoU loss
Correct answer: Mean squared error (MSE)
Original YOLO uses MSE for coordinate regression, though later versions switched to IoU-based losses.
Question 3: In SSD (Single Shot MultiBox Detector), what is the purpose of using multiple feature map scales?
- To reduce inference time
- To detect objects of different sizes (Correct answer)
- To increase classification accuracy
- To apply data augmentation
Correct answer: To detect objects of different sizes
SSD predicts detections from feature maps at multiple resolutions, allowing it to handle both small and large objects.
Question 4: What does 'anchor-free' mean in the context of object detectors like FCOS?
- No backbone network is used
- Bounding boxes are predicted without predefined anchor boxes (Correct answer)
- No classification head is used
- No NMS post-processing is applied
Correct answer: Bounding boxes are predicted without predefined anchor boxes
Anchor-free detectors like FCOS predict object locations directly from feature map points without relying on predefined anchor boxes.
Question 5: Which metric is typically used to evaluate object detection performance?
- Top-5 accuracy
- Mean Average Precision (mAP) (Correct answer)
- F1 score
- Pixel accuracy
Correct answer: Mean Average Precision (mAP)
mAP averages the Average Precision across all object classes and IoU thresholds, making it the standard detection metric.
Question 6: In the Feature Pyramid Network (FPN), how are features from different levels combined?
- By concatenation only
- By top-down pathway with lateral connections (Correct answer)
- By average pooling across levels
- By a transformer attention mechanism
Correct answer: By top-down pathway with lateral connections
FPN uses a top-down pathway that upsamples high-level features and adds them to lower-level features via lateral connections.
Question 7: What problem does Non-Maximum Suppression (NMS) solve in object detection?
- Underfitting of the classification head
- Multiple overlapping predictions for the same object (Correct answer)
- Gradient vanishing in deep networks
- Class imbalance during training
Correct answer: Multiple overlapping predictions for the same object
NMS removes redundant bounding boxes by keeping only the highest-scoring box among overlapping predictions for the same object.
CNN Object Detection Models 2
What is the primary role of the Region Proposal Network (RPN) in Faster R-CNN?