Research map / Video / Video segmentation and tracking
Time Lens++: Event-Based Frame Interpolation With Parametric Non-Linear Flow and Multi-Scale Fusion
Stepan Tulyakov, Alfredo Bochicchio, Daniel Gehrig, Stamatios Georgoulis, Yuanyou Li, Davide Scaramuzza
Check 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 |
|---|---|---|---|
| replace linear motion with cubic spline flow (ablation) | Time Lens++ motion module · Vimeo90k validation | PSNR [dB]: 26.83 -> 27.41 | Improved |
| replace spline motion with non-parametric flow (ablation) | Time Lens++ motion module · Vimeo90k validation | PSNR [dB]: 27.41 -> 28.20 | Improved |
| replace spline motion with non-parametric flow (ablation) | Time Lens++ motion module · 15 flow predictions | runtime [ms]: 220 -> 2700 | Hurt |
| add images to event-based motion estimation (ablation) | Time Lens++ motion module · Vimeo90k validation | PSNR [dB]: 26.82 -> 27.41 | Improved |
| add events to image-based motion estimation (ablation) | Time Lens++ motion module · Vimeo90k validation | PSNR [dB]: 24.51 -> 27.41 | Improved |
| combine synthesis and warping features (ablation) | Time Lens++ fusion module · Vimeo90k validation | PSNR [dB]: 29.77 -> 31.87 | Improved |
| add gated compression to feature fusion (ablation) | Time Lens++ fusion module · Vimeo90k validation | PSNR [dB]: 31.67 -> 31.87 | Improved |
| use separate image and event encoders instead of one U-Net (ablation) | Time Lens++ motion module · Vimeo90k validation | PSNR [dB]: 27.13 -> 27.41 | Improved |
| use multistage instead of single-stage training (ablation) | Time Lens++ motion module · Vimeo90k validation | PSNR [dB]: 27.41 -> 28.10 | Improved |
| share weights between the two synthesis encoders (ablation) | Time Lens++ fusion module · Vimeo90k validation | PSNR [dB]: 31.74 -> 31.87 | Improved |
| use multistage instead of single-stage training (ablation) | Time Lens++ fusion module · Vimeo90k validation | PSNR [dB]: 31.87 -> 32.73 | Improved |
| replace Time Lens fusion with multi-scale feature-level fusion | Time Lens++ · Vimeo90k validation | PSNR [dB]: 31.25 -> 32.73 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Time Lens — Time Lens: Event-Based Video Frame Interpolation (CVPR 2021) · better on 5 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Bridge Frame and Event: Common Spatiotemporal Fusion for High-Dynamic Scene Optical Flow (CVPR 2025 · similarity 0.88)
- Video Frame Interpolation via Direct Synthesis with the Event-based Reference (CVPR 2024 · similarity 0.88)
- TimeTracker: Event-based Continuous Point Tracking for Video Frame Interpolation with Non-linear Motion (CVPR 2025 · similarity 0.88)
- Repurposing Pre-trained Video Diffusion Models for Event-based Video Interpolation (CVPR 2025 · similarity 0.87)
- Injecting Frame-Event Complementary Fusion into Diffusion for Optical Flow in Challenging Scenes (NeurIPS 2025 · similarity 0.87)
- TTA-EVF: Test-Time Adaptation for Event-based Video Frame Interpolation via Reliable Pixel and Sample Estimation (CVPR 2024 · 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.
