I am analyzing the performance of an autonomous vehicle’s pedestrian detection system, and I want to ensure that I am interpreting the scenario correctly in terms of confusion-matrix components. This question is different from similar examples because I am specifically trying to resolve an ambiguity in the problem statement, not simply calculate the metrics.
Here is the situation:
“An autonomous vehicle's pedestrian detection system is being tested. Out of 150 pedestrian detections, 20 were missed (actual pedestrians not detected), and 15 false alarms (detecting a pedestrian where there was none) occurred.”
My goal is to determine how this information should be mapped to the four confusion-matrix categories:
- True positives (TP)
- False positives (FP)
- False negatives (FN)
- True negatives (TN)
I also intend to compute common evaluation metrics, including accuracy, precision, recall, specificity, and F1 score.
