Our hope is that this stimulus set can be of use to researchers interested in mid-level feature representations. It is now openly available on OSF: osf.io/7c9bz/. Feel free to reach out with questions or suggestions! (3/3).
Using this stimulus set + encoding models, we showed that 1) mid-level features are represented between ~100-250 ms in humans and 2) the mid-level feature processing hierarchy in humans is mirrored in CNNs. (2/3).
New preprint! www.biorxiv.org/content/10.1... We created and put to the test a new stimulus set of naturalistic scene images and videos, along with their ground-truth annotations for five mid-level features. (1/3)
