I run evaluate.py to load MSLS val (sf, cph) for evaluate predictions. The number of database images loaded are 18871. But only 747 query images are loaded.
This is obviously lower than the number of actual images of queries in val set. I wonder if this is correct.
I run evaluate.py to load MSLS val (sf, cph) for evaluate predictions. The number of database images loaded are 18871. But only 747 query images are loaded.
This is obviously lower than the number of actual images of queries in val set. I wonder if this is correct.