This happened on the validation dataset.
I am training the yolov8n-nano backbone model for object detection, and for 2 of my classes out of the 42 classes I am training, I get a precision of 100% while recall, mAP50 and mAP50-95 remain at 0%
I re-checked those classes by plotting predictions on the image along with ground truth and found no detection for those classes, yet the result was 100% precision and 0% recall.
i expect that if there is no detection all the metrics should be 0%