Research map / 3D vision / Depth and stereo
Taming Contrast Maximization for Learning Sequential, Low-latency, Event-based Optical Flow
Federico Paredes-Vallés, Kirk Y. W. Scheper, Christophe De Wagter, Guido C. H. E. de Croon
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 |
|---|---|---|---|
| process event stream sequentially with short windows and recurrent model (ablation) | contrast-maximization optical flow · DSEC-Flow | EPE: 3.48 -> 2.33 | Improved |
| set training partition length R=10 instead of 2, 5 or 20 (ablation) | dt=0.01s model · DSEC-Flow | EPE: R=2 9.66, R=5 4.05, R=10 2.33, R=20 16.63 | Mixed |
| add multi-timescale loss (S>1) (ablation) | dt=0.01s model · DSEC-Flow | EPE: 2.33 -> 2.82 (R=10,S=3), 2.73 (R=20,S=4) | Mixed |
| add multi-timescale loss (S>1) when training partition is long (ablation) | dt=0.01s, R=20 · DSEC-Flow | EPE: 16.63 -> 2.73 | Improved |
| shorten input window dt from 0.01s to 0.002s (ablation) | sequential model · DSEC-Flow | EPE: 2.33 -> 2.34 (0.005s), 2.66 (0.002s) | No clear effect |
| use the proposed event warping and loss augmentations over prior sequential contrast maximization | ConvGRU-EV-FlowNet pipeline · DSEC-Flow | EPE: 4.27 -> 2.33 | Improved |
| rely on FWL deblurring metric to judge flow quality | evaluation metrics · DSEC-Flow | FWL: FWL 1.91 at EPE 9.66 | Hurt |
| transfer DSEC-trained model to MVSEC without retraining | dt=0.005s, R=20 model · MVSEC outdoor_day1 | EPE: 0.27 vs 0.47 (ConvGRU-EV-FlowNet SSL) | Improved |
Closest papers by meaning
- Motion-prior Contrast Maximization for Dense Continuous-Time Motion Estimation (ECCV 2024 · similarity 0.91)
- Secrets of Event-Based Optical Flow (ECCV 2022 · similarity 0.91)
- Back to Event Basics: Self-Supervised Learning of Image Reconstruction for Event Cameras via Photometric Constancy (CVPR 2021 · similarity 0.90)
- Event Cameras, Contrast Maximization and Reward Functions: An Analysis (CVPR 2019 · similarity 0.90)
- Secrets of Event-Based Optical Flow, Depth and Ego-Motion Estimation by Contrast Maximization (TPAMI 2024 · similarity 0.89)
- Graph Neural Network Combining Event Stream and Periodic Aggregation for Low-Latency Event-based Vision (CVPR 2025 · similarity 0.87)
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.
