Research map / 3D vision / Depth and stereo
DeepTAM: Deep Tracking and Mapping
Huizhong Zhou, Benjamin Ummenhofer, Thomas Brox
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 learned optical flow features to tracking network (ablation) | DeepTAM tracking · RGB-D benchmark average | translational RMSE (m/s): 0.050 -> 0.043 | Improved |
| predict multiple pose hypotheses instead of a single pose (ablation) | DeepTAM tracking · RGB-D benchmark average | translational RMSE (m/s): 0.043 -> 0.040 | Improved |
| learned tracking versus classic RGB-D SLAM keyframe odometry | DeepTAM tracking · RGB-D benchmark average | translational RMSE (m/s): 0.060 -> 0.040 | Improved |
| combined DeepTAM tracking and mapping versus CNN-SLAM | DeepTAM · RGB-D benchmark average | translational RMSE (m/s): 0.253 -> 0.086 | Improved |
| use 6 frames instead of 2 in the fixed band cost volume (ablation) | DeepTAM mapping · MVS test | sc-inv: 0.193 -> 0.160 | Improved |
| use 10 frames instead of 6 in the fixed band cost volume (ablation) | DeepTAM mapping · MVS test | sc-inv: 0.160 -> 0.159 | No clear effect |
| add narrow band refinement, 3 iterations (ablation) | DeepTAM mapping · MVS test | sc-inv: 0.156 -> 0.132 | Improved |
| increase narrow band iterations from 3 to 5 (ablation) | DeepTAM mapping · MVS test | sc-inv: 0.132 -> 0.130 | No clear effect |
| increase narrow band iterations from 3 to 5 (ablation) | DeepTAM mapping · SUN3D test | sc-inv: 0.125 -> 0.128 | Hurt |
| DeepTAM mapping versus DTAM on weakly textured indoor data | DeepTAM mapping · SUNCG test | sc-inv: 0.343 -> 0.128 | Improved |
| use 64 labels instead of 32 for SGM and DTAM cost volume (ablation) | SGM, DTAM baselines · mapping comparison | depth error | Improved |
| add noise to camera poses | SGM, DTAM vs DeepTAM · mapping with noisy poses | depth quality | Mixed |
Later papers that compared with this method
3 later accepted papers report DeepTAM as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- DiffPoseNet: Direct Differentiable Camera Pose Estimation (CVPR 2022 · better on 1 of 2 shared results)
- Tangent Space Backpropagation for 3D Transformation Groups (CVPR 2021 · better on 1 of 1 shared result)
- DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Cameras (NeurIPS 2021 · better on 1 of 1 shared result)
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.
- DeMoN — DeMoN: Depth and Motion Network for Learning Monocular Stereo (CVPR 2017) · 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
- Dense Dynamic Scene Reconstruction and Camera Pose Estimation from Multi-View Videos (ECCV 2026 · similarity 0.87)
- KV-Tracker: Real-Time Pose Tracking with Transformers (CVPR 2026 · similarity 0.86)
- CNN-SLAM: Real-Time Dense Monocular SLAM With Learned Depth Prediction (CVPR 2017 · similarity 0.86)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · similarity 0.85)
- DELTA: DENSE EFFICIENT LONG-RANGE 3D TRACKING FOR ANY VIDEO (ICLR 2025 · similarity 0.85)
- MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos (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.
