CNN Object Detection Models 3

Question 1

How does RetinaNet address the foreground-background class imbalance problem?

Accepted Answer

By applying Focal Loss

Answer

RetinaNet uses Focal Loss, which down-weights the contribution of easy negatives so training focuses on hard examples.

Question 2

What is the key architectural difference between one-stage and two-stage object detectors?

Accepted Answer

Two-stage detectors generate region proposals first, then classify them

Answer

Two-stage detectors (e.g., Faster R-CNN) first generate proposals then classify each; one-stage detectors (e.g., YOLO) predict boxes and classes simultaneously.

Question 3

In YOLO, what does each grid cell predict?

Accepted Answer

Bounding boxes, objectness scores, and class probabilities

Answer

Each YOLO grid cell predicts B bounding boxes with confidence scores and C class probabilities for objects whose center falls in that cell.

Question 4

What is Intersection over Union (IoU) used for in object detection?

Accepted Answer

Quantifying the overlap between predicted and ground-truth boxes

Answer

IoU measures the ratio of overlap area to union area between a predicted box and a ground-truth box, indicating detection quality.

Question 5

Which component of Faster R-CNN is shared between the RPN and the detection head?

Accepted Answer

The convolutional feature extractor (backbone)

Answer

Both the RPN and the detection head use the same convolutional backbone features, making computation efficient.

CNN - Convolutional Neural Networks Practice Test

CNN - Convolutional Neural Networks Practice Test

CNN Object Detection Models 3