Research map / 3D vision / Face and shape
RANA: Relightable Articulated Neural Avatars
Umar Iqbal, Akin Caliskan, Koki Nagano, Sameh Khamis, Pavlo Molchanov, Jan Kautz
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 |
|---|---|---|---|
| add albedo regularization loss (ablation) | RANA · Relighting Humans protocol (a) | image LPIPS: 0.249 -> 0.217 | Improved |
| add coarse geometry and albedo texture inputs (ablation) | RANA · Relighting Humans protocol (a) | image LPIPS: 0.242 -> 0.217 | Improved |
| pretrain AlbedoNet and NormalNet on 400 synthetic characters (ablation) | RANA · Relighting Humans protocol (a) | image LPIPS: 0.301 -> 0.217 | Improved |
| pretrain on synthetic data (ablation) | RANA · Relighting Humans protocol (a) | normal error (degrees): 74.215 -> 64.350 | Improved |
| also fine-tune NormalNet during personalization (ablation) | RANA · Relighting Humans protocol (a) | image LPIPS: 0.217 -> 0.219 | No clear effect |
| use RANA instead of SMPL+D baseline for albedo maps | RANA · Relighting Humans protocol (a), albedo | albedo LPIPS: 0.216 (SMPL+D) vs 0.219 (RANA) | Hurt |
| RANA versus Relighting4D for novel light synthesis | RANA · Relighting Humans protocol (b) | image LPIPS: 0.192 -> 0.173 | Improved |
| RANA with pretraining needs 15k iterations versus 260k for Relighting4D | RANA · training cost | training iterations: 260k -> 15k | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Relighting4D — Relighting4D: Neural Relightable Human from Videos (ECCV 2022) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Relightable Neural Actor with Intrinsic Decomposition and Pose Control (ECCV 2024 · similarity 0.90)
- Relightable and Animatable Neural Avatars from Videos (AAAI 2024 · similarity 0.90)
- NECA: Neural Customizable Human Avatar (CVPR 2024 · similarity 0.88)
- Artist-Friendly Relightable and Animatable Neural Heads (CVPR 2024 · similarity 0.88)
- IntrinsicAvatar: Physically Based Inverse Rendering of Dynamic Humans from Monocular Videos via Explicit Ray Tracing (CVPR 2024 · similarity 0.87)
- RAM-Avatar: Real-time Photo-Realistic Avatar from Monocular Videos with Full-body Control (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.
