Research map / Video / Video segmentation and tracking
Learned video compression
Replaces traditional codecs with end-to-end networks using motion compensation, feature-space prediction and transformers.
68 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2018–2026) · an approach in Video segmentation and tracking, keyword cluster “compression · coding · rate”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
16 papers in 2023–24 and 30 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 3.6% to 5.6%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2018 · Multi-Frame Quality Enhancement for Compressed Video (CVPR 2018 · 218 citations)
- 2019 · DVC: An End-To-End Deep Video Compression Framework (CVPR 2019 · 744 citations)
- 2020 · Scale-Space Flow for End-to-End Optimized Video Compression (CVPR 2020 · 316 citations)
- 2021 · FVC: A New Framework Towards Deep Video Compression in Feature Space (CVPR 2021 · 273 citations)
- 2022 · Coarse-To-Fine Deep Video Coding With Hyperprior-Guided Mode Prediction (CVPR 2022 · 106 citations)
- 2023 · Neural Video Compression With Diverse Contexts (CVPR 2023 · 246 citations)
- 2024 · Neural Video Compression with Feature Modulation (CVPR 2024 · 140 citations)
- 2025 · Towards Practical Real-Time Neural Video Compression (CVPR 2025 · 61 citations)
- 2026 · Turbo-VAED: Fast and Stable Transfer of Video-VAEs to Mobile Devices (AAAI 2026 · 2 citations)
Earliest papers here
- Multi-Frame Quality Enhancement for Compressed Video (CVPR 2018 · 218 citations)
- Disentangled Sequential Autoencoder (ICML 2018 · 87 citations)
- Learning Binary Residual Representations for Domain-Specific Video Streaming (AAAI 2018 · 36 citations)
What has been tried
51 papers here report 489 measured changes (a component added, removed or swapped, compared with a baseline): helped 78%, no clear effect 3%, hurt 17%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add structure optimization | All frames, YUV420, intra-period –1 · BD-Rate (%) | 0.8 | Hurt | Neural Video Compression with Feature Modulation (CVPR 2024 · 140 citations) |
| use implicit temporal modelling instead of explicit motion modules | MCL-JCV, large motion · BD-Rate | 0.0% -> 2.1% | Hurt | Towards Practical Real-Time Neural Video Compression (CVPR 2025 · 61 citations) |
| Add quantization step < 1 | Kodak · BD-rate vs C3 Adaptive | 2.6% | Helped | C3: High-Performance and Low-Complexity Neural Compression from a Single Image or Video (CVPR 2024 · 50 citations) |
| Add offline Spynet fine-tuning with VTM motion vectors | All testing datasets (Average) · BD-Rate (%) | -3.9 | Helped | Offline and Online Optical Flow Enhancement for Deep Video Compression (AAAI 2024 · 26 citations) |
| add temporal probabilistic prediction | UVG · PSNR | Helped | Deep Hierarchical Video Compression (AAAI 2024 · 25 citations) | |
| Add motion vectors to the coding-prior-free baseline | JCT-VC, LDB, QP 37, average of 18 sequences · ∆PSNR (dB) | 0.69 -> 0.72 | Helped | CPGA: Coding Priors-Guided Aggregation Network for Compressed Video Quality Enhancement (CVPR 2024 · 21 citations) |
| add time-dependent feature stream ft | Bartender · PSNR | 30.59 -> 31.94 | Helped | GIFStream: 4D Gaussian-based Immersive Video with Feature Stream (CVPR 2025 · 14 citations) |
| Add MNLC to the DCVC-DC base | HEVC B/C/D/E, IP 32 · BD-Rate (%) vs base model | −4.0 / −9.9 / −9.0 / −9.7 | Helped | ECVC: Exploiting Non-Local Correlations in Multiple Frames for Contextual Video Compression (CVPR 2025 · 13 citations) |
| Add flow orientation | HEVC datasets, intra-period 32 · BD-Rate (%) | 0.0 -> -1.9 | Helped | Neural Video Compression with Context Modulation (CVPR 2025 · 13 citations) |
| Use DivGoP with Pm frames instead of Pr frames | MOT17 · mAP | 28.68% vs 32.67% bitrate savings | Helped | Task-Aware Encoder Control for Deep Video Compression (CVPR 2024 · 13 citations) |
| add motion estimation | N3DV · BD-PSNR | -1.86 dB | Helped | 4DGC: Rate-Aware 4D Gaussian Compression for Efficient Streamable Free-Viewpoint Video (CVPR 2025 · 12 citations) |
| Add tri-plane residual dynamic modeling | Volumetric video · bitrate | Helped | VRVVC: Variable-Rate NeRF-Based Volumetric Video Compression (AAAI 2025 · 6 citations) |
What reviewers will expect
Datasets and baselines named in the 16 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- Vimeo-90k · 7 of 16 papers
- N3DV · 2 of 16 papers
Baselines
- DCVC-DC · 7 of 16 papers
- DCVC-FM · 6 of 16 papers
- DCVC-TCM · 4 of 16 papers
- K-Planes · 3 of 16 papers
- ReRF · 3 of 16 papers
- DCVC-HEM · 3 of 16 papers
- TeTriRF · 2 of 16 papers
- NeRFPlayer · 2 of 16 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 46 papers; from OpenAlex and the papers' affiliations).
- University of Science and Technology of China · 6 papers
- Shanghai Jiao Tong University · 5 papers
- Microsoft Research Asia · 4 papers
- Nanjing University · 2 papers
- Hong Kong University of Science and Technology · 2 papers
Use it
Download BibTeX of 37 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Neural video representations (21 papers)
Elsewhere in the topic
- Self-supervised video representation learning (167 papers)
- Video restoration with temporal priors (159 papers)
- Layered video decomposition and dynamics (140 papers)
- Stochastic video prediction (125 papers)
- Physics-inspired motion prediction (113 papers)
- Sign language and physiological video analysis (99 papers)
- Weakly supervised video anomaly detection (99 papers)
- Spatio-temporal predictive modeling (91 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).