Research map / Video / Video segmentation and tracking
High Speed and High Dynamic Range Video with an Event Camera
Henri Rebecq, René Ranftl, Vladlen Koltun, Davide Scaramuzza
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 the learned recurrent network instead of HF and MR | event-to-video reconstruction · Event Camera Dataset | SSIM: 0.5009 -> 0.6157 | Improved |
| use the learned network instead of HF and MR | event-to-video reconstruction · Event Camera Dataset | LPIPS: 0.5173 -> 0.4092 | Improved |
| use the learned network instead of HF and MR on synthetic events | event-to-video reconstruction · Synthetic event sequences | LPIPS: 0.467 -> 0.302 | Improved |
| add the temporal loss (ablation) | E2VID · Event Camera Dataset | temporal error: 2.0810 -> 1.4257 | Improved |
| add the temporal loss (ablation) | E2VID · Event Camera Dataset | SSIM: 0.5935 -> 0.6157 | Improved |
| add the recurrent connection (ablation) | E2VID · Event Camera Dataset | temporal error: 3.9964 -> 1.4257 | Improved |
| add the recurrent connection (ablation) | E2VID · Event Camera Dataset | SSIM: 0.5364 -> 0.6157 | Improved |
| replace vanilla RNN with stacked ConvLSTM (ablation) | E2VID · tau=5 ms windows | SSIM: 0.2685 -> 0.5800 | Improved |
| replace vanilla RNN with stacked ConvLSTM (ablation) | E2VID · tau=50 ms windows | SSIM: 0.6100 -> 0.6157 | No clear effect |
| apply reconstructions to classification with a pretrained CNN instead of HATS | object classification · N-Caltech101 | accuracy: 0.700 -> 0.866 | Improved |
| apply reconstructions instead of HATS on a synthetic dataset | object classification · N-MNIST | accuracy: 0.991 -> 0.983 | Hurt |
| feed reconstructions to a CNN trained on image data without fine-tuning | object classification · N-Caltech101 | accuracy: 0.700 -> 0.821 | Improved |
Closest papers by meaning
- Events-To-Video: Bringing Modern Computer Vision to Event Cameras (CVPR 2019 · similarity 0.94)
- Learning To Reconstruct High Speed and High Dynamic Range Videos From Events (CVPR 2021 · similarity 0.92)
- Video to Events: Recycling Video Datasets for Event Cameras (CVPR 2020 · similarity 0.92)
- Simultaneous Optical Flow and Intensity Estimation From an Event Camera (CVPR 2016 · similarity 0.91)
- Joint Framework for Single Image Reconstruction and Super-Resolution With an Event Camera (TPAMI 2021 · similarity 0.91)
- Event-3DGS: Event-based 3D Reconstruction Using 3D Gaussian Splatting (NeurIPS 2024 · similarity 0.89)
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.
