Research map / Video / Video segmentation and tracking
Ultra-Fast Neural Video Compression
Jiahao Li, Wenxuan Xie, Zhaoyang Jia, Bin Li, Zongyu Guo, Xiaoyi Zhang, Yan Lu
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 chunk coding without frame-specific decoders (ablation) | DCVC-RT · UVG/HEVC/MCL-JCV average | BD-Rate and decoding FPS: -21.0% -> -10.1%; 105.3 -> 349.1 FPS | Mixed |
| add frame-specific decoders to chunk coding (ablation) | DCVC-UF chunk coding · average BD-Rate | BD-Rate: -10.1% -> -25.3% | Improved |
| add streamlined entropy model (ablation) | DCVC-UF chunk coding with frame-specific decoders · average BD-Rate and decoding FPS | decoding FPS: 343.2 -> 453.3 FPS; -25.3% -> -23.4% | Mixed |
| fine-tune with 128-frame training videos (ablation) | DCVC-UF (HT-S) · average BD-Rate | BD-Rate: -23.4% -> -31.6% | Improved |
| use larger network (HT-L) instead of HT-S (ablation) | DCVC-UF · average BD-Rate and 1080p 4090 decoding FPS | BD-Rate: -31.6% -> -42.2%; 453.3 -> 273.6 FPS | Mixed |
| use chunk size 8 instead of 1 frame (ablation) | DCVC-UF · average BD-Rate | BD-Rate: -9.5% -> -31.6% | Improved |
| reduce VTM hierarchical-B GOP size from 32 to 8 (ablation) | VTM-17.0 · average bitrate saving over VTM LD | bitrate saving: 33.8% -> 23.7% | Hurt |
Closest papers by meaning
- Towards Practical Real-Time Neural Video Compression (CVPR 2025 · similarity 0.93)
- Neural Video Compression with Feature Modulation (CVPR 2024 · similarity 0.90)
- Neural Video Compression with Context Modulation (CVPR 2025 · similarity 0.90)
- NVRC: Neural Video Representation Compression (NeurIPS 2024 · similarity 0.90)
- PNVC: Towards Practical INR-based Video Compression (AAAI 2025 · similarity 0.89)
- C3: High-Performance and Low-Complexity Neural Compression from a Single Image or Video (CVPR 2024 · 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.
