Research map / 3D vision / Face and shape
Texture Fields: Learning Texture Representations in Function Space
Michael Oechsle, Lars Mescheder, Michael Niemeyer, Thilo Strauss, Andreas Geiger
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use Texture Fields instead of the projection baseline (GT shapes) | single-image texture reconstruction · ShapeNet single-image texture reconstruction (GT shapes), mean | FID: 25.936 -> 12.036 | Improved |
| use Texture Fields instead of the NVS baseline (GT shapes) | single-image texture reconstruction · ShapeNet single-image texture reconstruction (GT shapes), mean | FID: 27.031 -> 12.036 | Improved |
| use Texture Fields instead of NVS (GT shapes) | single-image texture reconstruction · ShapeNet single-image texture reconstruction (GT shapes), mean | SSIM: 0.938 -> 0.934 | Hurt |
| use Texture Fields instead of NVS (GT shapes) | single-image texture reconstruction · ShapeNet single-image texture reconstruction (GT shapes), mean | Feature-l1: 0.159 -> 0.144 | Improved |
| use Texture Fields instead of NVS in the full pipeline | Occupancy Networks shape + texture · ShapeNet full pipeline with Occupancy Networks shapes, mean | FID: 69.306 -> 51.659 | Improved |
| use Texture Fields instead of NVS in the full pipeline | Occupancy Networks shape + texture · ShapeNet full pipeline with Occupancy Networks shapes, mean | SSIM: 0.867 -> 0.867 | No clear effect |
| use Texture Fields instead of NVS in the full pipeline | Occupancy Networks shape + texture · ShapeNet full pipeline with Occupancy Networks shapes, mean | Feature-l1: 0.225 -> 0.211 | Improved |
| use Texture Fields instead of Im2Avatar (cars, full pipeline) | textured 3D reconstruction · ShapeNet full pipeline with Occupancy Networks shapes, cars | FID: 149.393 -> 77.439 | Improved |
| use Texture Fields instead of projection on tables (GT shapes) | single-image texture reconstruction · ShapeNet single-image texture reconstruction (GT shapes), tables | Feature-l1: 0.123 -> 0.123 | No clear effect |
| use continuous texture field instead of 128^3 voxel representation | cat/human single-object fitting · single 3D object overfitting | high-frequency detail (qualitative) | Improved |
Closest papers by meaning
- Convolutional Generation of Textured 3D Meshes (NeurIPS 2020 · similarity 0.89)
- Fine Detailed Texture Learning for 3D Meshes With Generative Models (TPAMI 2023 · similarity 0.89)
- Mesh2Tex: Generating Mesh Textures from Image Queries (ICCV 2023 · similarity 0.89)
- Leveraging 2D Data to Learn Textured 3D Mesh Generation (CVPR 2020 · similarity 0.89)
- TUVF: Learning Generalizable Texture UV Radiance Fields (ICLR 2024 · similarity 0.88)
- Single Mesh Diffusion Models with Field Latents for Texture Generation (CVPR 2024 · similarity 0.87)
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.
