Research map / Video / Video generation
NeRV-Diffusion: Diffuse Implicit Neural Representation for Video Synthesis
Yixuan Ren, Hanyu Wang, Bo He, Hao Chen, Abhinav Shrivastava
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 channel-wise INR parameterization instead of repeat modulation (ablation) | NeRV autoencoder (S) · UCF-101 | gFVD: 741 -> 570 | Improved |
| Add multi-head affine mapping to reuse weight tokens across layers (ablation) | NeRV autoencoder (S) · UCF-101 | gFVD: 570 -> 283 | Improved |
| Direct reuse of tokens without multi-head affines (ablation) | NeRV autoencoder (S) · UCF-101 | gFVD: 570 -> 562 | No clear effect |
| Use 8x8 spatiotemporal input embedding instead of 1x1 (ablation) | Generative NeRV (S) · UCF-101 | gFVD: 283 -> 254 | Improved |
| Use 16x16 spatial embedding rather than 8x8 (ablation) | Generative NeRV (S) · UCF-101 | gFVD: 254 -> 277 | Hurt |
| Use transposed convolution instead of PixelShuffle (ablation) | Generative NeRV (S) · UCF-101 | gFVD: 254 -> 248 | Improved |
| Add residual side connections (ablation) | Generative NeRV (S) · UCF-101 | gFVD: 248 -> 219 | Improved |
| Increase number of weight tokens from 128 to 256 (ablation) | NeRV-Diffusion-S · UCF-101 | gFVD: 184 -> 206 | Hurt |
| Use DiT denoiser instead of Latte (temporal attention) (ablation) | Implicit diffusion · UCF-101 | gFVD: 342 -> 295 | Improved |
| Add Min-SNR-gamma loss weighting (ablation) | Implicit DiT · UCF-101 | gFVD: 295 -> 238 | Improved |
| Add scheduled sampling (ablation) | Implicit DiT · UCF-101 | gFVD: 295 -> 261 | Improved |
| Combine Min-SNR-gamma and scheduled sampling (ablation) | Implicit DiT · UCF-101 | gFVD: 295 -> 184 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- CogVideo — CogVideo: Large-scale Pretraining for Text-to-Video Generation via Transformers (ICLR 2023) · better on 2 of 2 shared results
- VideoFusion — VideoFusion: Decomposed Diffusion Models for High-Quality Video Generation (CVPR 2023) · better on 1 of 1 shared result
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
- Efficient Video Diffusion Models via Content-Frame Motion-Latent Decomposition (ICLR 2024 · similarity 0.89)
- DS-NeRV: Implicit Neural Video Representation with Decomposed Static and Dynamic Codes (CVPR 2024 · similarity 0.89)
- Combining Frame and GOP Embeddings for Neural Video Representation (CVPR 2024 · similarity 0.88)
- Boosting Neural Representations for Videos with a Conditional Decoder (CVPR 2024 · similarity 0.88)
- OSV: One Step is Enough for High-Quality Image to Video Generation (CVPR 2025 · similarity 0.88)
- HyperNVD: Accelerating Neural Video Decomposition via Hypernetworks (CVPR 2025 · similarity 0.88)
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.
