Research map / Label-efficient and robust learning / Self-supervised and contrastive
TESPEC: Temporally-Enhanced Self-Supervised Pretraining for Event Cameras
Mohammad Mohammadi, Ziyi Wu, Igor Gilitschenski
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 masking (50%) versus autoencoding with 0% masking (ablation) | TESPEC · DSEC semantic segmentation | mIoU: 55.619 -> 62.774 | Improved |
| raise masking ratio from 25% to 50% (ablation) | TESPEC · DSEC semantic segmentation | mIoU: 61.625 -> 62.774 | Improved |
| raise masking ratio from 50% to 75% (ablation) | TESPEC · DSEC semantic segmentation | mIoU: 62.774 -> 61.812 | Hurt |
| reconstruct improved intensity video instead of event histograms (ablation) | TESPEC · DSEC semantic segmentation | mIoU: 60.430 -> 62.774 | Improved |
| use proposed global intensity estimation instead of prior per-pixel estimation (ablation) | TESPEC · DSEC semantic segmentation | mIoU: 60.493 -> 62.774 | Improved |
| use proposed global intensity estimation instead of prior per-pixel estimation (ablation) | TESPEC · DDD17 semantic segmentation | mIoU: 63.835 -> 65.187 | Improved |
| set normalization factor N to 5,000 instead of 200,000 (ablation) | TESPEC · DSEC semantic segmentation | mIoU: 60.059 -> 62.774 | Improved |
| set normalization factor N to 5,000 instead of 500 (ablation) | TESPEC · DSEC semantic segmentation | mIoU: 61.742 -> 62.774 | Improved |
| pre-train on 1Mpx instead of Gen1 (ablation) | TESPEC · DSEC semantic segmentation | mIoU: 60.128 -> 62.774 | Improved |
| compute MSE on normalized patches (ablation) | TESPEC · DSEC semantic segmentation | mIoU: 61.792 -> 62.774 | Improved |
| use MAE objective instead of CPC (ablation) | TESPEC · DSEC semantic segmentation | mIoU: 59.921 -> 62.774 | Improved |
| use MAE objective instead of ECDDP contrastive loss (ablation) | TESPEC · DSEC semantic segmentation | mIoU: 61.012 -> 62.774 | Improved |
Closest papers by meaning
- Efficient Event Camera Data Pretraining with Adaptive Prompt Fusion (ICCV 2025 · similarity 0.87)
- Event Camera Data Dense Pre-training (ECCV 2024 · similarity 0.87)
- TETO: Tracking Events with Teacher Observation for Motion Estimation and Frame Interpolation (ECCV 2026 · similarity 0.86)
- No More Shortcuts: Realizing the Potential of Temporal Self-Supervision (AAAI 2024 · similarity 0.86)
- PASS: Path-selective State Space Model for Event-based Recognition (NeurIPS 2025 · similarity 0.86)
- ESEG: Event-Based Segmentation Boosted by Explicit Edge-Semantic Guidance (AAAI 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.
