TALK DETAILS VIDEO RECORDING DOI QR CODE RELATED TALKS SEARCH POSTERS

TALK DETAILS

Mooney Face Image Processing in Deep Convolutional Neural Networks Compared to Humans

Astrid Zeman — Tim Leers, Hans Op de Beeck

Show Affils
First Author
► Astrid Zeman — The University of Melbourne

Contributors
► Tim Leers — KU Leuven
► Hans Op de Beeck — KU Leuven
28 September 2022
Deep Convolutional Neural Networks (CNNs) are criticised for their reliance on local shape features and texture rather than global shape. We test whether CNNs are able to process global shape information in the absence of local shape cues and texture by testing their performance on Mooney stimuli, which are face images thresholded to binary values. More specifically, we assess whether CNNs classify these abstract stimuli as face-like, and whether they exhibit the face inversion effect (FIE), where upright stimuli are classified positively at a higher rate compared to inverted. We tested a CNN trained for facial recognition (DeepFace) and found that the network performs perceptual completion and exhibits the FIE, which is present over all levels of specificity. By matching the false positive rate of CNNs to humans, we found that the network performed closer to the human average (85.73% for upright, 57.25% for inverted) for both conditions (62.70% for upright, 42.26% for inverted). Rank order correlation between DeepFace and humans across individual stimuli shows a small but significant correlation in upright and inverted conditions, indicating a relationship in image difficulty between observers and the model. We conclude that despite texture and local shape bias of CNNs, which makes their performance distinct from humans, they are still able to process object images holistically.
doi.org/10.57736/nmc-d07a-e915📋

VIDEO RECORDING

QR CODE

TALKS YOU MIGHT BE INTERESTED IN

📃 Differential representation of natural and manmade images in the human ventral visual stream
Mrugsen Nagsen Gopnarayan — Deeksha Rathore, Fabio Bauer, Jasper Hilliard, Prerita Chawla, Raffe Sharif
📃 Comparing CNNs and the brain: sensitivity to images altered in the frequency domain
Alexander Claman — Xu Pan, Vanessa Aguiar-Pulido, Odelia Schwartz
📃 Saccade Mechanisms for Image Classification, Object Detection and Tracking
Zachary Daniels — Saurabh Farkya, Zachary Daniels, Aswin Nadamuni Raghavan, David Zhang, Michael Piacentino