Research map / 3D vision / 3D generation
AC3S: Adaptive Conditioning for 3D-Aware Synthetic Data Generation
Eric Ji, Qiran Hu, Wufei Ma, Sarthak Jain, Yingying Li, Minh Do, Yaoyao Liu
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 visual prompt modulator (ablation) | 3D-DST baseline · FID, 50 ImageNet classes (Overall) | FID: 87.19 -> 77.53 | Improved |
| add multi-agent VLM text prompt generator (ablation) | 3D-DST baseline · FID, 50 ImageNet classes (Overall) | FID: 87.19 -> 78.93 | Improved |
| combine modulator and VLM prompts (ablation) | 3D-DST baseline · FID, 50 ImageNet classes (Overall) | FID: 87.19 -> 71.24 | Improved |
| combine modulator and VLM prompts (ablation) | 3D-DST baseline · FID, animal classes | FID: 72.67 -> 48.87 | Improved |
| add VLM prompt generator on top of modulator alone (objects) (ablation) | AC3S · FID, object classes | FID: 81.18 -> 76.15 | Improved |
| pretrain on AC3S instead of 3D-DST data | ConvNeXt-S · ImageNet top-1 overall | top-1 accuracy: 81.89 -> 82.75 | Improved |
| pretrain on AC3S instead of 3D-DST data | Swin-S · ImageNet top-1 overall | top-1 accuracy: 83.17 -> 84.19 | Improved |
| pretrain on AC3S instead of 3D-DST data | ResNet-18 · ImageNet top-1 overall | top-1 accuracy: 76.96 -> 77.33 | Improved |
| pretrain on Text2Img data instead of none | ConvNeXt-S · ImageNet top-1 overall | top-1 accuracy: 80.12 -> 79.21 | Hurt |
| pretrain on AC3S data instead of none | ConvNeXt-S and Swin-S · ImageNet top-1 overall | top-1 accuracy: 80.12 -> 82.75; 81.59 -> 84.19 | Improved |
| pretrain on 3D-DST instead of AC3S | ConvNeXt-S · ImageNet top-1 animals | top-1 accuracy: 95.39 -> 95.14 | Hurt |
| pretrain on AC3S instead of 3D-DST | ResNet-18 · PASCAL3D+ Sofa | Acc@pi/18: 37.83 -> 38.92 | Improved |
Closest papers by meaning
- Ouroboros3D: Image-to-3D Generation via 3D-aware Recursive Diffusion (CVPR 2025 · similarity 0.88)
- Generating Images with 3D Annotations Using Diffusion Models (ICLR 2024 · similarity 0.88)
- CAD: Photorealistic 3D Generation via Adversarial Distillation (CVPR 2024 · similarity 0.88)
- Sculpt3D: Multi-View Consistent Text-to-3D Generation with Sparse 3D Prior (CVPR 2024 · similarity 0.88)
- Efficient-3Dim: Learning a Generalizable Single-image Novel-view Synthesizer in One Day (ICLR 2024 · similarity 0.87)
- AC3D: Analyzing and Improving 3D Camera Control in Video Diffusion Transformers (CVPR 2025 · 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.
