Research map / Architectures and efficiency / RNNs and normalization
Deep Patch Visual Odometry
Zachary Teed, Lahav Lipson, Jia Deng
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 sparse patch-based correspondence instead of dense flow (DPVO vs DROID-VO) | DPVO · EuRoC MAV | ATE [m] avg: 0.186 -> 0.105 | Improved |
| use sparse patch-based correspondence instead of dense flow (DPVO vs DROID-VO) | DPVO · TartanAir test split | ATE avg: 0.58 -> 0.21 | Improved |
| use sparse patch-based correspondence instead of dense flow (DPVO vs DROID-VO) | DPVO · TUM-RGBD freiburg1 | ATE avg: 0.098 -> 0.089 | Improved |
| use 96 patches / 10-frame window (Default) instead of 48 patches / 7-frame (Fast) | DPVO · EuRoC MAV | ATE [m] avg: 0.129 -> 0.105 | Improved |
| use 96 patches / 10-frame window (Default) instead of Fast on ICL-NUIM | DPVO · ICL-NUIM | ATE avg: 0.093 -> 0.097 | No clear effect |
| use patches (larger than 1x1) instead of point features (ablation) | DPVO · TartanAir validation | accuracy | Improved |
| add 1D temporal convolution to update operator (ablation) | DPVO · TartanAir validation | accuracy | Improved |
| add softmax aggregation (message passing) to update operator (ablation) | DPVO · TartanAir validation | accuracy | Improved |
| random patch centroids instead of SIFT/ORB/SuperPoint/high-gradient keypoints (ablation) | DPVO · TartanAir validation | accuracy | Improved |
| increase number of patches beyond 96 (ablation) | DPVO · EuRoC | ATE | No clear effect |
Closest papers by meaning
- Deep Patch Visual SLAM (ECCV 2024 · similarity 0.90)
- Leveraging Consistent Spatio-Temporal Correspondence for Robust Visual Odometry (AAAI 2025 · similarity 0.89)
- ZeroVO: Visual Odometry with Minimal Assumptions (CVPR 2025 · similarity 0.86)
- LEAP-VO: Long-term Effective Any Point Tracking for Visual Odometry (CVPR 2024 · similarity 0.86)
- Adaptive VIO: Deep Visual-Inertial Odometry with Online Continual Learning (CVPR 2024 · similarity 0.86)
- MambaVO: Deep Visual Odometry Based on Sequential Matching Refinement and Training Smoothing (CVPR 2025 · similarity 0.85)
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.
