I was curious how they obtained both the list of source images and the file names, etc for the source images (and whether that together was a one-step process or a two step process)
I did have to do some amount of flickr scraping from the dataset. The browser also aligns and crops the faces according the way nvidia did as well. I thought it would be an interesting commentary to look at the faces in the same way that the machine did.
6 comments
[ 3.5 ms ] story [ 25.1 ms ] thread