Research map / 3D vision / NeRF and radiance fields
Unconstrained Scene Generation With Locally Conditioned Radiance Fields
Terrance DeVries, Miguel Angel Bautista, Nitish Srivastava, Graham W. Taylor, Joshua M. Susskind
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 local latent grid conditioning instead of a global latent code (ablation) | GSN · Replica 64x64 | FID: 68.42 -> 41.75 | Improved |
| add local latent grid conditioning instead of a global latent code (ablation) | GSN · Replica 64x64 | SwAV-FID: 6.14 -> 4.14 | Improved |
| use local coordinate system for local radiance fields instead of global (ablation) | GSN · Replica 64x64 latent grid rotation | FID | Improved |
| train with long camera trajectories instead of short ones (ablation) | GSN · Replica 64x64 | FID on long trajectories | Improved |
| add depth input to the discriminator (any resolution) versus no depth (ablation) | GSN · Replica 64x64 | FID: 311.35 -> 41.75 | Improved |
| reduce depth resolution from 32 to a single pixel (ablation) | GSN · Replica 64x64 | FID: 41.75 -> 41.57 | No clear effect |
| reduce depth resolution from 32 to 4 (ablation) | GSN · Replica 64x64 | FID: 41.75 -> 44.18 | No clear effect |
| add R1 gradient penalty to the discriminator (ablation) | GSN · Replica 64x64 | FID: 52.9 -> 41.75 | Improved |
| add discriminator reconstruction penalty (ablation) | GSN · Replica 64x64 | FID: 274.3 -> 41.75 | Improved |
| add data augmentation to discriminator training (ablation) | GSN · Replica 64x64 | FID: 412.25 -> 41.75 | Improved |
| use GSN with locally conditioned radiance field instead of GRAF | GRAF · VizDoom 64x64 | FID: 47.50 -> 37.21 | Improved |
| use GSN instead of GRAF | GRAF · Replica 64x64 | FID: 65.37 -> 41.75 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- pi-GAN — Pi-GAN: Periodic Implicit Generative Adversarial Networks for 3D-Aware Image Synthesis (CVPR 2021) · better on 6 of 6 shared results
- GRAF — GRAF: Generative Radiance Fields for 3D-Aware Image Synthesis (NeurIPS 2020) · better on 6 of 6 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
- GRF: Learning a General Radiance Field for 3D Representation and Rendering (ICCV 2021 · similarity 0.90)
- GRAF: Generative Radiance Fields for 3D-Aware Image Synthesis (NeurIPS 2020 · similarity 0.90)
- SparseGNV: Generating Novel Views of Indoor Scenes with Sparse RGB-D Images (AAAI 2024 · similarity 0.89)
- NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis (ECCV 2020 · similarity 0.89)
- LOLNerf: Learn From One Look (CVPR 2022 · similarity 0.89)
- DisCoScene: Spatially Disentangled Generative Radiance Fields for Controllable 3D-Aware Scene Synthesis (CVPR 2023 · similarity 0.88)
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.
