Research map / Label-efficient and robust learning / Self-supervised and contrastive
Event Camera Data Pre-training
Yan Yang, Liyuan Pan, Liu Liu
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 |
|---|---|---|---|
| use event-specific data augmentations applied before event image conversion (ablation) | Event pre-training (ViT-S/16) · N-ImageNet linear probing, 100 epochs | top-1 accuracy: 46.1% -> 53.5% | Improved |
| scale backbone from ViT-S/16 to ViT-L/16 (ablation) | Event pre-training · N-ImageNet fine-tuning | top-1 accuracy: 64.84 -> 71.05 | Improved |
| use proposed method instead of MAE pre-training on event data | ViT backbones · N-ImageNet fine-tuning | top-1 accuracy: 55.45 -> 64.83 | Improved |
| pre-train with proposed framework vs training ViT-S/16 from scratch | ViT-S/16 · N-ImageNet | top-1 accuracy: 46.70 -> 64.83 | Improved |
| apply RGB-image-image SSL methods (iBoT, BeiT, MoCo-v3) directly to event data | Existing SSL methods · N-ImageNet fine-tuning | top-1 accuracy: iBoT 19.55; MoCo-v3 45.77 vs ViT-S/16 scratch 46.70 | Hurt |
| pre-train with proposed framework on semantic segmentation (ResNet50) instead of ViT-S/16 | Event pre-training · DSEC | mIoU: 47.91 -> 59.16 | Improved |
| pre-train ResNet50 with proposed method vs training from scratch | ResNet50 · DSEC | mIoU: 57.60 -> 59.16 | Improved |
| pre-train ViT-S/16 with proposed method vs scratch | ViT-S/16 · MVSEC indoor_flying1 | AEE: 0.68 -> 0.61 | Improved |
| pre-train ViT-S/16 with proposed method vs scratch | ViT-S/16 · DDD17 | mIoU: 48.76 -> 54.66 | Improved |
Closest papers by meaning
- Event Camera Data Dense Pre-training (ECCV 2024 · similarity 0.91)
- Events-To-Video: Bringing Modern Computer Vision to Event Cameras (CVPR 2019 · similarity 0.87)
- Efficient Event Camera Data Pretraining with Adaptive Prompt Fusion (ICCV 2025 · similarity 0.87)
- Back to Event Basics: Self-Supervised Learning of Image Reconstruction for Event Cameras via Photometric Constancy (CVPR 2021 · similarity 0.87)
- Video to Events: Recycling Video Datasets for Event Cameras (CVPR 2020 · similarity 0.87)
- EZSR: Event-based Zero-Shot Recognition (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.
