Research map / 3D vision / 3D generation
Simple and Effective Synthesis of Indoor 3D Scenes
Jing Yu Koh, Harsh Agrawal, Dhruv Batra, Richard Tucker, Austin Waters, Honglak Lee, Yinfei Yang, Jason Baldridge, Peter Anderson
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 the direct guidance-image-to-RGB-D GAN instead of Pathdreamer | Indoor novel-view synthesis on Matterport3D · Matterport3D Val-Unseen, 1 context, 1 step | FID: 27.2 -> 19.6 | Improved |
| Use the direct guidance-image-to-RGB-D GAN instead of Pathdreamer | Indoor novel-view synthesis on Matterport3D · Matterport3D Val-Unseen, 1 context, 1–6 steps | FID: 65.8 -> 58.0 | Improved |
| Use the direct guidance-image-to-RGB-D GAN instead of PixelSynth | Novel-view synthesis · RealEstate10K val, all images | FID: 25.5 -> 23.5 | Improved |
| Remove L1 depth reconstruction loss (ablation) | Ours · Matterport3D Val-Unseen, 1 context, 1 step | FID: 19.6 -> 22.7 | Hurt |
| Add VGG perceptual loss (ablation) | Ours · Matterport3D Val-Unseen, 1 context | FID: FID@1: 19.6 -> 18.2; FID@{1-6}: 58.0 -> 66.5 | Mixed |
| Add KLD loss for noise conditioning (ablation) | Ours · Matterport3D Val-Unseen, 1 context | FID: FID@1: 19.6 -> 22.6; FID@{1-6}: 58.0 -> 65.2 | Hurt |
| Use one shared decoder instead of separate RGB and depth decoders (ablation) | Ours · Matterport3D Val-Unseen, 1 context | FID: FID@1: 19.6 -> 24.1; FID@{1-6}: 58.0 -> 56.9 | Mixed |
| Add random guidance masking and four additional convolutional layers (ablation) | Ours · Matterport3D Val-Unseen, 1 context | FID: FID@1: 24.2 -> 19.6; FID@{1-6}: 68.0 -> 58.0 | Improved |
| Replace generated depth with ground-truth depth in the GAN loss (ablation) | Ours · Matterport3D Val-Unseen, 1 context | FID: FID@1: 19.6 -> 24.3; FID@{1-6}: 58.0 -> 66.1 | Hurt |
| Add synthetic trajectory augmentation using Ours | VLN-BERT (MURAL) · R2R Val-Unseen | Success Rate: 67.3 -> 68.1 | Improved |
| Add Habitat renders alongside Ours synthetic trajectory augmentation | VLN-BERT (MURAL) · R2R Val-Unseen | Success Rate: 67.3 -> 68.8; +1.5% | Improved |
| Add SimCLR image-based augmentation | VLN-BERT (MURAL) · R2R Val-Unseen | Success Rate: 67.3 -> 66.5 | Hurt |
Closest papers by meaning
- SceneWeaver: All-in-One 3D Scene Synthesis with an Extensible and Self-Reflective Agent (NeurIPS 2025 · similarity 0.88)
- SparseGNV: Generating Novel Views of Indoor Scenes with Sparse RGB-D Images (AAAI 2024 · similarity 0.88)
- InstructScene: Instruction-Driven 3D Indoor Scene Synthesis with Semantic Graph Prior (ICLR 2024 · similarity 0.88)
- InternScenes: A Large-scale Simulatable Indoor Scene Dataset with Realistic Layouts (NeurIPS 2025 · similarity 0.88)
- SceneCraft: Layout-Guided 3D Scene Generation (NeurIPS 2024 · similarity 0.88)
- CAD: Photorealistic 3D Generation via Adversarial Distillation (CVPR 2024 · 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.
