Research map / Detection and segmentation / Open-set and few-shot detection
Generating Furry Cars: Disentangling Object Shape and Appearance across Multiple Domains
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| add learned feature histograms and the hybrid appearance-consistency objective | FineGAN · birds ↔ cars | color histogram χ²-distance: 0.4295 -> 0.2484 | Helped |
| add learned feature histograms and the hybrid appearance-consistency objective | FineGAN · cars ↔ dogs | texture histogram χ²-distance: 0.4006 -> 0.2535 | Helped |
| add learned feature histograms and the hybrid appearance-consistency objective | FineGAN · dogs ↔ birds | color histogram χ²-distance: 0.2862 -> 0.2009 | Helped |
| add learned feature histograms and the hybrid appearance-consistency objective | FineGAN · animals ↔ animals | texture histogram χ²-distance: 0.2322 -> 0.0695 | Helped |
| learn the feature filters instead of using a fixed random filter bank (ablation) | Ours · birds ↔ cars | color histogram χ²-distance: 0.2842 -> 0.2484 | Helped |
| learn the feature filters instead of using a fixed random filter bank (ablation) | Ours · cars ↔ dogs | texture histogram χ²-distance: 0.2754 -> 0.2535 | Helped |
| learn the feature filters instead of using a fixed random filter bank (ablation) | Ours · animals ↔ animals | color histogram χ²-distance: 0.0721 -> 0.0462 | Helped |
| relax the hierarchical constraint between shape and appearance codes (ablation) | FineGAN · birds ↔ cars | color histogram χ²-distance: 0.4295 -> 0.3121 | Hurt |
| use filters with increasing receptive field instead of a single 1x1 filter layer (ablation) | Ours · animals ↔ animals | texture appearance transfer | Helped |
| use three 3x3 filter layers instead of two 3x3 filter layers (ablation) | Ours · animals ↔ animals | appearance transfer | No clear effect |
| optimize both the shape and appearance stages rather than only the appearance stage (ablation) | Ours · animals ↔ animals | appearance transfer | Helped |
| vary the contrastive-loss temperature among 0.1, 0.5, and 0.9 (ablation) | Ours · birds ↔ cars | appearance transfer | No clear effect |
Closest papers by meaning
- Unsupervised Part-Based Disentangling of Object Shape and Appearance (CVPR 2019 · similarity 0.89)
- Visual Object Networks: Image Generation with Disentangled 3D Representations (NeurIPS 2018 · similarity 0.88)
- Explicit Disentanglement of Appearance and Perspective in Generative Models (NeurIPS 2019 · similarity 0.88)
- Disentangled Pose and Appearance Guidance for Multi-Pose Generation (CVPR 2025 · similarity 0.85)
- Disentangled 3D Scene Generation with Layout Learning (ICML 2024 · similarity 0.85)
- Diffusion 3D Features (Diff3F): Decorating Untextured Shapes with Distilled Semantic Features (CVPR 2024 · similarity 0.85)
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.