CNN Object Detection Models 5 — Questions and Answers
Question 1: What distinguishes instance segmentation from semantic segmentation in the context of object detection?
- Instance segmentation only works on single-object images
- Instance segmentation distinguishes between separate instances of the same class (Correct answer)
- Semantic segmentation provides bounding boxes while instance segmentation does not
- Instance segmentation uses only fully connected layers
Correct answer: Instance segmentation distinguishes between separate instances of the same class
Instance segmentation assigns a unique mask to each individual object instance, while semantic segmentation labels each pixel with a class without distinguishing instances.
Question 2: In the context of YOLO training, what is the purpose of using k-means clustering on the training dataset?
- To cluster images by scene complexity
- To determine good anchor box dimensions for the dataset (Correct answer)
- To group similar classes together
- To balance the number of positive and negative examples
Correct answer: To determine good anchor box dimensions for the dataset
K-means clustering on ground-truth bounding box dimensions helps choose anchor sizes that best fit the objects in the specific training dataset.
Question 3: What is the main trade-off when increasing the number of anchor boxes per location in a detector?
- Higher recall but increased computational cost and memory (Correct answer)
- Lower recall but faster inference
- Better generalization but worse localization
- Faster training but lower mAP
Correct answer: Higher recall but increased computational cost and memory
More anchors improve coverage of object shapes and sizes (higher recall) but increase the number of predictions to process, raising computation and memory cost.
Question 4: Which of the following best describes the 'neck' component in modern object detection architectures like YOLOv4?
- The final classification head
- Feature aggregation module between backbone and detection head (Correct answer)
- The input preprocessing pipeline
- The loss computation layer
Correct answer: Feature aggregation module between backbone and detection head
The neck (e.g., PANet, FPN) aggregates and mixes features from different backbone stages before passing them to the detection head.
Question 5: Why does soft-NMS outperform standard NMS in crowded scene detection?
- It removes all overlapping boxes regardless of score
- It decays scores of overlapping boxes rather than eliminating them (Correct answer)
- It uses a learned threshold instead of a fixed IoU cutoff
- It applies NMS only to the highest confidence class
Correct answer: It decays scores of overlapping boxes rather than eliminating them
Soft-NMS reduces suppressed boxes' scores by a continuous function of overlap instead of hard removal, preserving detections of nearby objects.
Question 6: What is the 'class imbalance problem' that one-stage detectors face during training?
- Some object classes are harder to classify than others
- There are vastly more background anchor boxes than foreground object anchors (Correct answer)
- The model learns faster for large objects than small ones
- Different feature map scales have unequal numbers of anchor boxes
Correct answer: There are vastly more background anchor boxes than foreground object anchors
In one-stage detectors, the majority of anchor boxes cover background, creating a massive imbalance that can dominate the loss and degrade training.
Question 7: In COCO evaluation, what does AP@[0.5:0.95] represent?
- Accuracy at a fixed IoU threshold of 0.5
- Average Precision averaged over IoU thresholds from 0.5 to 0.95 in 0.05 steps (Correct answer)
- The precision at 95% recall
- mAP computed only on objects with size between 0.5 and 0.95 of image area
Correct answer: Average Precision averaged over IoU thresholds from 0.5 to 0.95 in 0.05 steps
COCO's primary metric averages AP across ten IoU thresholds (0.50, 0.55, …, 0.95), rewarding more precise localization than the VOC AP@0.5 metric.
CNN Object Detection Models 5
What distinguishes instance segmentation from semantic segmentation in the context of object detection?