Research map / 3D vision / Gaussian splatting
RealLiFe: Real-Time Light Field Reconstruction via Hierarchical Sparse Gradient Descent
Yijie Deng, Lei Han, Tianpeng Lin, Lin Li, Jinzhi Zhang, Lu Fang
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 sparsification factor k=5 instead of k=3 (ablation) | RealLiFe · Real Forward-Facing (ablation) | PSNR: 24.95 -> 25.46 | Improved |
| increase k from 5 to 7 (ablation) | RealLiFe · Real Forward-Facing (ablation) | PSNR: 25.46 -> 25.51 | No clear effect |
| increase k from 5 to 7 (ablation) | RealLiFe · Real Forward-Facing (ablation) | Generation FPS: 37.04 -> 32.79 | Hurt |
| use 40 MPI planes instead of 16 (ablation) | RealLiFe · Real Forward-Facing (ablation) | PSNR: 24.39 -> 25.46 | Improved |
| increase planes from 40 to 64 (ablation) | RealLiFe · Real Forward-Facing (ablation) | PSNR: 25.46 -> 25.48 | No clear effect |
| increase planes from 40 to 64 (ablation) | RealLiFe · Real Forward-Facing (ablation) | Generation FPS: 37.04 -> 28.57 | Hurt |
| use 3 iterations instead of 2 (ablation) | RealLiFe · Real Forward-Facing (ablation) | PSNR: 25.33 -> 25.46 | No clear effect |
| use 3 iterations instead of 2 (ablation) | RealLiFe · Real Forward-Facing (ablation) | Generation FPS: 45.45 -> 37.04 | Hurt |
| add sparsity loss (ablation) | RealLiFe · Real Forward-Facing (ablation) | PSNR: 25.20 -> 25.46 | Improved |
| add sparsity loss (qualitative example) (ablation) | RealLiFe · Qualitative example image | PSNR: 27.42 -> 28.51 | Improved |
| use 3D CNN backbone instead of MLP (ablation) | RealLiFe · Real Forward-Facing (ablation) | PSNR: 24.86 -> 25.46 | Improved |
| use alpha-gradient top-k sparsification instead of MVS-sampling (ablation) | RealLiFe · Real Forward-Facing (ablation) | PSNR: 24.34 -> 25.46 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- DeepView — DeepView: View Synthesis With Learned Gradient Descent (CVPR 2019) · better on 0 of 3 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
- A Generative Model for Generic Light Field Reconstruction (TPAMI 2020 · similarity 0.88)
- S2D-LFE: Sparse-to-Dense Light Field Event Generation (CVPR 2025 · similarity 0.88)
- SparseCam4D: Spatio-Temporally Consistent 4D Reconstruction from Sparse Cameras (CVPR 2026 · similarity 0.87)
- Direct Voxel Grid Optimization: Super-Fast Convergence for Radiance Fields Reconstruction (CVPR 2022 · similarity 0.87)
- Light Field Neural Rendering (CVPR 2022 · similarity 0.87)
- ZeroRF: Fast Sparse View 360deg Reconstruction with Zero Pretraining (CVPR 2024 · similarity 0.86)
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.
