Research map / Video / Video segmentation and tracking
Efficient Neural Supersampling on a Novel Gaming Dataset
Antoine Mercier, Ruan Erasmus, Yashesh Savani, Manik Dhingra, Fatih Porikli, Guillaume Berger
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 depth-informed motion vector dilation (ablation) | Ours-M (f32-m3) · Gaming dataset test scenes (2x) | PSNR: 31.97 -> 31.80 | Hurt |
| add blending with previous output (ablation) | Ours-M (f32-m3) · AbandonedSchool static frames | pixel-wise std: 1.10 -> 0.54 | Improved |
| add jittered input sampling (ablation) | Ours-M (f32-m3) · AbandonedSchool static frames | PSNR: 35.06 -> 37.38 | Improved |
| use dataset motion vectors instead of RAFT-estimated flow (ablation) | Ours-M (f32-m3) · Gaming dataset test scenes (2x) | PSNR: 31.09 -> 31.80 | Improved |
| add warping (ablation) | Ours-M (f32-m3) · Gaming dataset test scenes (2x) | PSNR: 30.69 -> 31.80 | Improved |
| use multiple frames instead of single frame (ablation) | Ours-M (f32-m3) · Gaming dataset test scenes (2x) | PSNR: 30.56 -> 31.80 | Improved |
| add jitter-conditioned kernels in first and last convs (ablation) | Ours-M (f32-m3) · Gaming dataset test scenes (2x) | PSNR: 31.26 -> 31.80 | Improved |
| add jitter-conditioned kernel in the last conv only (ablation) | Ours-M (f32-m3) · Gaming dataset test scenes (2x) | PSNR: 31.50 -> 31.80 | Improved |
| add jitter-conditioned kernel in the first conv only (ablation) | Ours-M (f32-m3) · Gaming dataset test scenes (2x) | PSNR: 31.68 -> 31.80 | Improved |
| train on all 10 scenes instead of dropping the last 5 (ablation) | Ours-M (f32-m3) · AbandonedSchool static frames | pixel-wise std: 0.76 -> 0.54 | Improved |
| train on all 10 scenes instead of dropping 8 (ablation) | Ours-M (f32-m3) · Gaming dataset test scenes (2x) | PSNR: 31.56 -> 31.80 | Improved |
| replace L1 loss with SSIM plus VGG perceptual loss (ablation) | Ours-M (f32-m3) · Gaming dataset test scenes (2x) | LPIPS: 0.140 -> 0.125 | Improved |
Closest papers by meaning
- Low-Latency Space-Time Supersampling for Real-Time Rendering (AAAI 2024 · similarity 0.87)
- Achieving Lightweight Super-Resolution for Real-Time Computer Graphics (AAAI 2025 · similarity 0.87)
- Neural Super-Resolution for Real-time Rendering with Radiance Demodulation (CVPR 2024 · similarity 0.86)
- Efficient Video Super-Resolution for Real-time Rendering with Decoupled G-buffer Guidance (CVPR 2025 · similarity 0.85)
- High-Quality Real-Time Rendering Using Subpixel Sampling Reconstruction (AAAI 2024 · similarity 0.84)
- Video2Game: Real-time Interactive Realistic and Browser-Compatible Environment from a Single Video (CVPR 2024 · similarity 0.84)
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.
