AI Paper Atlas

Research map / 3D vision / Face and shape

Fabric Material Recovery from Video Using Multi-Scale Geometric Auto-Encoder

Junbang Liang, Ming Lin

ECCV 2022 · 3D vision › Face and shape · approach: Clothed human modeling · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use LSTM temporal modelling instead of single-frame CNN, image features (ablation)Image-only model · Fabric material classificationmean accuracy (%): 5.11% -> 45.27%Improved
add garment latent features to image features (LSTM) (ablation)Image-only LSTM · Fabric material classificationmean accuracy (%): 45.27% -> 70.14%Improved
add image features to garment-only features (LSTM) (ablation)Garment-only LSTM · Fabric material classificationmean accuracy (%): 65.16% -> 70.14%Improved
add LSTM instead of CNN on image+garment features (ablation)Image + Garment · Fabric material classificationmean accuracy (%): 12.62% -> 70.14%Improved
replace AtlasNet with the two-level auto-encoder (ablation)Garment auto-encoder · Ablation (auto-encoder)material accuracy (%): 54.20% -> 70.14%Improved
replace parallel separate estimation with joint closed-loop body-garment estimation (ablation)Single-frame estimator · Ablation (human and garment estimation)material accuracy (%): 49.15% -> 70.14%Improved
use learned model instead of optimization-simulation baselineMaterial estimation · Mid-day/Sunset test scenesaccuracy (%) and speed: T-shirt 80.2 -> 86.5, Skirt 83.3 -> 81.6; 4-6 hours -> 8.7 secMixed
use our garment model instead of MGN without fine-tuningGarment reconstruction · Multi-Garment Net datasetreconstruction error (mm): T-shirt 11.97 -> 3.08Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy