Research map / 3D vision / Depth and stereo
SENSE: A Shared Encoder Network for Scene-Flow Estimation
Huaizu Jiang, Deqing Sun, Varun Jampani, Zhaoyang Lv, Erik Learned-Miller, Jan Kautz
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 flow and segmentation tasks to a shared encoder (ablation) | SENSE disparity · KITTI 2015 40-image split | D1-occ: 2.73% -> 2.59% | Improved |
| add disparity and segmentation tasks to a shared encoder (ablation) | SENSE flow · KITTI 2015 40-image split | F1-occ: 11.37% -> 11.19% | Improved |
| add flow and disparity tasks to a shared encoder (ablation) | SENSE segmentation · KITTI 2015 40-image split | mIoU: 47.51% -> 48.25% | Improved |
| share one encoder between flow and disparity instead of separate encoders (ablation) | SENSE disparity · KITTI 2015 | D1-occ: 6.62% -> 6.38% | Improved |
| share one encoder between flow and disparity instead of separate encoders (ablation) | SENSE flow · KITTI 2015 | EPE vs F1-occ: EPE 6.37 -> 6.23; F1-occ 21.15% -> 23.29% | Mixed |
| use 5 pyramid levels and add Batch Normalization (ablation) | PWC-Net-style disparity decoder · FlyingThings3D | EPE: 2.10 -> 1.61 | Improved |
| replace plain encoder with ResNet residual blocks (ablation) | disparity decoder · FlyingThings3D | EPE: 1.61 -> 1.40 | Improved |
| add Pyramid Pooling Module (ablation) | disparity decoder · FlyingThings3D | EPE: 1.40 -> 1.32 | Improved |
| add hourglass refinement module (ablation) | disparity decoder · FlyingThings3D | EPE: 1.32 -> 1.15 | Improved |
| add segmentation distillation loss (ablation) | SENSE · KITTI 2015 40-image split | D1-occ: 2.52% -> 2.44% | Improved |
| add occlusion distillation loss (ablation) | SENSE · KITTI 2015 40-image split | D1-occ: 2.52% -> 2.38% | Improved |
| add semantic, photometric and structural self-supervised losses (ablation) | SENSE · KITTI 2015 40-image split | D1-occ: 2.52% -> 2.58% | No clear effect |
Closest papers by meaning
- Learning End-to-End Scene Flow by Distilling Single Tasks Knowledge (AAAI 2020 · similarity 0.90)
- Feature-Level Collaboration: Joint Unsupervised Learning of Optical Flow, Stereo Depth and Camera Motion (CVPR 2021 · similarity 0.90)
- Self-Supervised Monocular Scene Flow Estimation (CVPR 2020 · similarity 0.90)
- DeltaFlow: An Efficient Multi-frame Scene Flow Estimation Method (NeurIPS 2025 · similarity 0.88)
- Recurrent Partial Kernel Network for Efficient Optical Flow Estimation (AAAI 2024 · similarity 0.88)
- UnSAMFlow: Unsupervised Optical Flow Guided by Segment Anything Model (CVPR 2024 · similarity 0.88)
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.
