Research map / 3D vision / Depth and stereo
DAGE: Dual-Stream Architecture for Efficient and Fine-Grained Geometry Estimation
Tuan Duc Ngo, Jiahui Huang, Seoung Wug Oh, Kevin Blackburn-Matzen, Evangelos Kalogerakis, Chuang Gan, Joon-Young Lee
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 cross-attn + self-attn adapter after HR stream instead of aligned MoGe2 (post-align) (ablation) | PDTrack · NRGBD | Acc.: 0.031 -> 0.018 | Improved |
| use cross-attn + self-attn adapter instead of interpolation + self-attention (ablation) | PDTrack · NRGBD | Acc.: 0.030 -> 0.018 | Improved |
| use cross-attn + self-attn adapter instead of all cross-attention (ablation) | PDTrack · NRGBD | Acc.: 0.021 -> 0.018 | Improved |
| insert adapter after HR-stream ViT instead of alternating after last 5 blocks (ablation) | PDTrack · NRGBD | Acc.: 0.023 -> 0.018 | Improved |
| add monocular prior (ablation) | PDTrack · Sintel | F1: 0.27 -> 0.34 | Improved |
| add gradient loss (ablation) | PDTrack · Sintel | F1: 0.31 -> 0.34 | Improved |
| add local loss (ablation) | PDTrack · Sintel | F1: 0.34 -> 0.30 | Hurt |
| add local loss (ablation) | PDTrack · Sintel | Rel^p: 21.5 -> 20.9 | Improved |
| increase LR stream resolution from 252px to 518px (ablation) | PDTrack · Sintel depth | Rel^p: 21.5 -> 19.7 | Improved |
| increase LR stream resolution from 252px to 518px (ablation) | PDTrack · throughput | FPS: 65.4 -> 22.5 | Hurt |
| add distillation to LR stream (ablation) | PDTrack · Sintel pose | RPE_R: 0.584 -> 0.406 | Improved |
| add distillation to LR stream (ablation) | PDTrack · Sintel pose | ATE: 0.111 -> 0.132 | Hurt |
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.
- VGGT — VGGT: Visual Geometry Grounded Transformer (CVPR 2025) · better on 4 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
- FrameVGGT: Coherence-Preserving Memory for Bounded Streaming Geometry (NeurIPS 2026 · similarity 0.88)
- TurboVGGT: Fast Visual Geometry Reconstruction with Adaptive Alternating Attention (NeurIPS 2026 · similarity 0.87)
- MoGe-2: Accurate Monocular Geometry with Metric Scale and Sharp Details (NeurIPS 2025 · similarity 0.86)
- MoGe: Unlocking Accurate Monocular Geometry Estimation for Open-Domain Images with Optimal Training Supervision (CVPR 2025 · similarity 0.86)
- Not All Tasks Quantize Equally: Fisher-Guided Quantization for Visual Geometry Transformer (NeurIPS 2026 · similarity 0.86)
- World Tracing: Pixel-Aligned Geometry Beyond the Visible (NeurIPS 2026 · similarity 0.86)
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.
