Research map / Video / Video segmentation and tracking
VNVC: A Versatile Neural Video Coding Framework for Efficient Human-Machine Vision
Xihua Sheng, Li Li, Dong Liu, Houqiang Li
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 intermediate feature instead of reconstructed frame for temporal context (ablation) | VNVC · HEVC Class B | BD-rate (%): 2.2 -> 0.0 | Improved |
| use intermediate feature instead of last feature for temporal context (ablation) | VNVC · HEVC (average) | BD-rate (%): 0.74% BD-rate reduction | Improved |
| use intermediate feature instead of last feature for temporal context (ablation) | VNVC · HEVC Class E | BD-rate (%): 3.3 -> 0.0 | Improved |
| use intermediate feature instead of last feature for temporal context (ablation) | VNVC · decoding GFLOPs | GFLOPs: 6745.61 -> 4709.17 | Improved |
| use cross-domain motion coding instead of pixel-domain (ablation) | VNVC · encoding GFLOPs | GFLOPs: 7909.90 -> 4606.08 | Improved |
| use cross-domain motion coding instead of feature-domain (ablation) | VNVC · encoding time | seconds: 1.871 -> 1.336 | Improved |
| use cross-domain motion coding instead of pixel-domain (ablation) | VNVC · HEVC Class B | BD-rate (%): -1.6 -> 0.0 | Mixed |
| use 2 U-Net reconstruction network instead of 1 U-Net (ablation) | VNVC · HEVC Class E | BD-rate (%): 2.1 -> 0.0 | Improved |
| use 2 U-Net instead of 1 residual block (ablation) | VNVC · HEVC Class E | BD-rate (%): 6.5 -> 0.0 | Improved |
| decode intermediate features for the task (Ours(fea)) instead of decoded images (ablation) | VNVC · DAVIS 2017 segmentation | BD-performance: 0.00 vs -1.18 | Improved |
| decode intermediate features for the task (Ours(fea)) instead of decoded images (ablation) | VNVC · DAVIS decoding GFLOPs | GFLOPs: 1482.97 -> 1034.17 | Improved |
| decode intermediate features for the task instead of decoded images (ablation) | VNVC · DAVIS non-blind denoising (delta=20) | PSNR dB: 0.03 | No clear effect |
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.
- CANF-VC — CANF-VC: Conditional Augmented Normalizing Flows for Video Compression (ECCV 2022) · better on 3 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
- Video Coding for Machines: Compact Visual Representation Compression for Intelligent Collaborative Analytics (TPAMI 2024 · similarity 0.90)
- Towards Practical Real-Time Neural Video Compression (CVPR 2025 · similarity 0.88)
- Neural Video Compression with Feature Modulation (CVPR 2024 · similarity 0.88)
- NVRC: Neural Video Representation Compression (NeurIPS 2024 · similarity 0.88)
- PNVC: Towards Practical INR-based Video Compression (AAAI 2025 · similarity 0.87)
- Neural Video Compression with Context Modulation (CVPR 2025 · 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.
