Research map / 3D vision / Depth and stereo
Hyden: A Hybrid Dual-Path Encoder for Monocular Geometry of High-resolution Images
Zaiwei Zhang, Marc Mapeke, Wei Ye, Rakesh Ranjan, JQ Huang
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 crop loss to self-distillation (ablation) | Hyden-DA2 · iBims-1 boundary sharpness | F1: 11.8 -> 15.8 | Improved |
| increase local crops from 4 to 8 (ablation) | Hyden-DA2 · iBims-1/Sintel boundary sharpness | F1: iBims-1 15.8 -> 16.1; Sintel 33.1 -> 32.3 | No clear effect |
| train with mixed resolutions 518-2072 instead of 518-1036 (ablation) | Hyden-DA2 · NYUv2 relative depth | Rel: 5.14 -> 4.60 | Improved |
| use two-layer CNN fusion instead of MLP fusion (ablation) | Hyden-DA2 · HAMMER relative depth | Rel: 5.93 -> 5.44 | Improved |
| use two-layer CNN fusion instead of one-layer CNN (ablation) | Hyden-DA2 · HAMMER relative depth | Rel: 5.87 -> 5.44 | Improved |
| replace ViT encoder with hybrid dual-path encoder at 2K input | MoGe2 · inference latency | ms: 476.8 -> 171.6 | Improved |
| replace ViT encoder with hybrid dual-path encoder at 2K input | DepthAnything-v2 · inference latency | ms: 408.1 -> 100.7 | Improved |
| replace ViT encoder with hybrid encoder, low-resolution input 518 | MoGe2 / DA2 · depth accuracy at 518x518 | accuracy | Hurt |
| use Hyden self-distillation on MoGe2 | MoGe2 · zero-shot depth avg rank | rank: 2.0 -> 1.3 | Improved |
| use Hyden on MoGe2 for normals | MoGe2 · NYUv2 normal | mean angular error: 15.6 -> 14.6 | Improved |
| use Hyden on MoGe2 for metric depth without intrinsics | MoGe2 · NYUv2 metric depth | Rel: 9.2 -> 7.7 | Mixed |
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.
- MoGe2 — MoGe-2: Accurate Monocular Geometry with Metric Scale and Sharp Details (NeurIPS 2025) · better on 9 of 9 shared results
- DepthPro — Depth Pro: Sharp Monocular Metric Depth in Less Than a Second (ICLR 2025) · better on 3 of 3 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
- Multi-Resolution Monocular Depth Map Fusion by Self-Supervised Gradient-Based Composition (AAAI 2023 · similarity 0.88)
- Deep Digging into the Generalization of Self-Supervised Monocular Depth Estimation (AAAI 2023 · similarity 0.87)
- BRNet: Exploring Comprehensive Features for Monocular Depth Estimation (ECCV 2022 · similarity 0.87)
- HR-Depth: High Resolution Self-Supervised Monocular Depth Estimation (AAAI 2021 · similarity 0.87)
- Win-Win: Training High-Resolution Vision Transformers from Two Windows (ICLR 2024 · similarity 0.86)
- Improved Monocular Depth Prediction Using Distance Transform Over Pre-semantic Contours with Self-supervised Neural Networks (CVPR 2025 · similarity 0.84)
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.
