American University
Browse

AU Community Access Only

Reason: Restricted to American University authorized users. To access this content, please connect to the secure campus network (includes the AU VPN).

UNSUPERVISED LEARNING REVEALS INTERPRETABLE LATENT REPRESENTATION FOR TRANSLUCENCY PERCEPTION

thesis
posted on 2023-09-07, 02:01 authored by Chenxi Liao

Humans constantly assess the appearance of materials to plan actions, such as stepping on icy roads without slipping. Visual inference of materials is challenging because a given material can appear dramatically different in various scenes. This problem especially stands out for translucent materials (e.g., skin, soap), whose appearance strongly depends on lighting, geometry, and viewpoint. Despite this, humans can still distinguish between different materials, and it remains unsolved how to systematically discover visual features pertinent to material inference. Here, we developed an unsupervised style-based image generation model to identify perceptually relevant dimensions for material appearances from natural photographs. We find the model can synthesize images of convincing material appearances. Importantly, without supervision, human-understandable scene attributes, including object’s shape, material, and body color, spontaneously emerge in the model’s layer-wise latent space in a scale-specific manner. Crucially, the middle-layers of the latent space selectively encode translucency features correlated with perception, suggesting that translucent impressions are established in mid-to-low spatial scale features. Our findings indicate that humans may use the scale-specific statistical structure of natural images to efficiently represent material properties across contexts, and our approach is widely applicable in discovering perceptually relevant features from complex stimuli for many visual inference tasks.

History

Publisher

American University

Language

English

Committee chair

Bei Xiao

Committee member(s)

Zois Boukouvalas; Laurie Bayet; Arthur Shapiro

Degree discipline

Neuroscience

Degree grantor

American University. Department of Neuroscience

Degree level

  • Masters

Degree name

M.S. in Neuroscience

Local identifier

Liao_american_0008N_11942.pdf

Media type

application/pdf

Pagination

76 pages

Access statement

Electronic thesis unavailable until October 14, 2023, per author's request. Electronic thesis available to American University authorized users only, per author's request.

Submission ID

11942

Usage metrics

    Theses and Dissertations

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC