Image names from image indexes

Part 1 (2019)
1

Quick question:
How do I get the image names from the image indexes from the valid_ds?
I want to be able to identify the false positives/ false negatives, I have the indexes of the images, but I can’t find a way to recover the image name (to remove/modify them in the dataframe from which I created the ImageDataBunch).
I know these are the top losses images indexes from my validation ds:

 tensor([2118, 2025,  831,  164, 1426,  931, 2235, 1770, 1018, 2963, 2766, 1286,
         2914,  805, 1644, 1405, 2214, 1488,  377, 1768, 1587, 1509, 2321,  655,
         1861, 1483,  222, 1859,  915, 2633, 2525, 2459, 2275, 1081,  663, 1391,
         3022, 1722, 1073, 2693, 2406]))

How to get back the corresponding images from the df?

2

Ok, found that .items method gets the names back, for item 2118:

data.valid_ds.items[2118]
>>'./train_dehaze/img_107522017.jpg'