Is human object recognition invariant to depth cues?

Background. The human visual system can represent complex 3-D surfaces from multiple depth cues, but individual depth cues are processed by different cortical networks. For example, different cortical areas, scattered in the dorsal and ventral visual pathways, process depth from motion vs. shaded cues to extract 3-D surface information that may ultimately be used for object recognition. It is not known whether there is only one underlying mechanism for object recognition which combines multiple depth cues or whether there are multiple cue-specific representations of objects. Assuming efficiency in the object recognition mechanism, a single object representation is predicted to be formed in a specific brain region in the ventral visual pathway from multiple depth cues that are processed by different cortical areas. Thus, we were interested to investigate whether a complex object representation, such as a face representation, is achieved by a single depth cue and whether this representation is depth cue invariant by the level of the occipitotemporal cortex.Methods. We utilized the face identity aftereffect and the MEG adaptation paradigm to investigate the depth cue invariance. Face identity aftereffect is the transitory distortion in the perception of facial identity following extended periods of adaptation to a configurally distorted face called an anti-face. We measured face identification thresholds for each depth cue and across different depth cues in four subjects using a four-alternative forced-choice task with 3 conditions: matched anti-face adaptor, non-matched anti-face adaptor, as well as without adaptation. Using the MEG adaptation paradigm, we were interested to examine the adaptation of the face-selective M170 component, arising from the inferior occipitotemporal sources. We specifically asked whether this component showed reduced amplitude in response to a shaded face when it was preceded by a face, regardless of the depth cue defining the surface. We particularly looked at the adaptation of the face-selective M170 component across fourteen subjects and compared its amplitude and latency across different conditions.Results. We found robust face identity aftereffect not only from individual depth cues but also across different depth cues. The aftereffect was strongest in the shaded condition, weakest in the structure-from-motion condition, and comparable across texture and stereo disparity. Using the MEG adaptation paradigm, we found that part of the lateral occipital complex, particularly the right posterior inferior temporal gyrus, is tolerant to different depth cues. Conclusion. Despite the fact that our perception of the 3-D world is empowered by many depth cues, how depth cues contribute to object recognition is very poorly understood. Using the face identity aftereffect and the MEG adaptation approach we provided evidence for the existence of the depth cue invariance of complex object representations. These results are of fundamental importance to cortical organization of vision, but also valuable for developing object recognition algorithms for automated assessment based on 3-D information.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.