Research map / Video / Video segmentation and tracking
Layered video decomposition and dynamics
Decomposes videos into layers, objects and effects to model scene dynamics and generate videos.
140 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Video segmentation and tracking, keyword cluster “prediction · person · pose”
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.
19 papers in 2023–24 and 51 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 4.3% to 9.5%.
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
- 2016 · Face2Face: Real-Time Face Capture and Reenactment of RGB Videos (CVPR 2016 · 1,843 citations)
- 2017 · VidLoc: A Deep Spatio-Temporal Model for 6-DoF Video-Clip Relocalization (CVPR 2017 · 251 citations)
- 2018 · Gaze Prediction in Dynamic 360° Immersive Videos (CVPR 2018 · 283 citations)
- 2019 · Unsupervised Scale-consistent Depth and Ego-motion Learning from Monocular Video (NeurIPS 2019 · 123 citations)
- 2020 · Ego-Topo: Environment Affordances From Egocentric Video (CVPR 2020 · 99 citations)
- 2021 · VSPW: A Large-scale Dataset for Video Scene Parsing in the Wild (CVPR 2021 · 107 citations)
- 2022 · Deformable Sprites for Unsupervised Video Decomposition (CVPR 2022 · 47 citations)
- 2023 · HyperReel: High-Fidelity 6-DoF Video With Ray-Conditioned Sampling (CVPR 2023 · 124 citations)
- 2024 · CoDeF: Content Deformation Fields for Temporally Consistent Video Processing (CVPR 2024 · 48 citations)
- 2025 · Video Depth Anything: Consistent Depth Estimation for Super-Long Videos (CVPR 2025 · 61 citations)
- 2026 · VSCD: Video-based Scene Change Detection in Unaligned Scenes (ICML 2026 · 0 citations)
Earliest papers here
- Face2Face: Real-Time Face Capture and Reenactment of RGB Videos (CVPR 2016 · 1,843 citations)
- From Keyframes to Key Objects: Video Summarization by Representative Object Proposal Selection (CVPR 2016 · 128 citations)
- Walk and Learn: Facial Attribute Representation Learning From Egocentric Video and Contextual Data (CVPR 2016 · 96 citations)
What has been tried
111 papers here report 1,120 measured changes (a component added, removed or swapped, compared with a baseline): helped 74%, no clear effect 5%, hurt 19%, 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 |
|---|---|---|---|---|
| Use VideoAlign+SSI instead of VideoAlign | TartanAir and VKITTI · TAE | 1.326 -> 1.207 | Helped | Video Depth Anything: Consistent Depth Estimation for Super-Long Videos (CVPR 2025 · 61 citations) |
| replace positional encoding with 3D hash encoding | Video reconstruction · PSNR | +3.1 dB | Helped | CoDeF: Content Deformation Fields for Temporally Consistent Video Processing (CVPR 2024 · 48 citations) |
| Fine-tune the whole DUSt3R model rather than only the decoder | Sintel · Abs Rel | 0.306 -> 0.310 | Hurt | Align3R: Aligned Monocular Depth Estimation for Dynamic Videos (CVPR 2025 · 22 citations) |
| add BCC | DENSE-Fog light · Abs Rel | 0.636 -> 0.311 | Helped | Depth-Centric Dehazing and Depth-Estimation from Real-World Hazy Driving Video (AAAI 2025 · 15 citations) |
| add dilation rate 25 to snippet sampling | PointOdyssey · Abs Rel (%) | 16.7 -> 10.2 | Helped | Video Depth without Video Models (CVPR 2025 · 10 citations) |
| add cross-time attention | Egocentric Graph · accuracy | 82.00 -> 86.15 | Helped | The Audio-Visual Conversational Graph: From an Egocentric-Exocentric Perspective (CVPR 2024 · 10 citations) |
| add FIFA to the ResNet18-based method | IV · mean angular error | 8.05◦ -> 7.29◦ | Helped | FIFA: Fine-grained Inter-frame Attention for Driver's Video Gaze Estimation (CVPR 2025 · 9 citations) |
| Add local alignment loss | VMC ablation · Text-Align | 0.8163 -> 0.8382 | Helped | Spectral Motion Alignment for Video Motion Transfer Using Diffusion Models (AAAI 2025 · 8 citations) |
| Train on EgoExo4D rather than only the small AEA training set | ADT · PSNR | 15.49 -> 22.72 | Helped | 4DGT: Learning a 4D Gaussian Transformer Using Real-World Monocular Videos (NeurIPS 2025 · 8 citations) |
| Add LDM relationship modeling | Action Genome test, SGDET, w constraint · R@10 | 26.9 -> 29.7 | Helped | DiffVsgg: Diffusion-Driven Online Video Scene Graph Generation (CVPR 2025 · 6 citations) |
| add segmentation masks to distinguish foreground from background | DyCheck iPhone · mSSIM | 0.42 -> 0.60 | Helped | 4D-Fly: Fast 4D Reconstruction from a Single Monocular Video (CVPR 2025 · 5 citations) |
| Add global, time-invariant color mapping across frames to STG | Opera girl, Laboratory, and Puppy test views · PSNR | 28.482 -> 31.240; 26.306 -> 27.581; 20.497 -> 20.533 | Helped | ImViD: Immersive Volumetric Videos for Enhanced VR Engagement (CVPR 2025 · 5 citations) |
What reviewers will expect
Datasets and baselines named in the 36 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- DAVIS · 6 of 36 papers
- TartanAir · 4 of 36 papers
- ScanNet · 3 of 36 papers
- KITTI · 3 of 36 papers
- Sintel · 3 of 36 papers
- VKITTI · 2 of 36 papers
- PointOdyssey · 2 of 36 papers
- Bonn · 2 of 36 papers
Baselines
- DepthCrafter · 6 of 36 papers
- ChronoDepth · 5 of 36 papers
- Depth Anything V2 · 4 of 36 papers
- NVDS · 4 of 36 papers
- MonST3R · 3 of 36 papers
- Marigold · 3 of 36 papers
- Shape of Motion · 2 of 36 papers
- Depth Pro · 2 of 36 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 70 papers; from OpenAlex and the papers' affiliations).
- Zhejiang University · 4 papers
- Meta · 4 papers
- Hong Kong University of Science and Technology · 3 papers
- Stanford University · 3 papers
- Ant Group · 2 papers
- University of Chinese Academy of Sciences · 2 papers
- The Chinese University of Hong Kong · 2 papers
- Massachusetts Institute of Technology · 2 papers
- ByteDance · 2 papers
- The University of Hong Kong · 2 papers
Use it
Download BibTeX of 40 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
- Stochastic video prediction (125 papers)
- Physics-inspired motion prediction (113 papers)
- Spatial-temporal video person re-identification (41 papers)
Elsewhere in the topic
- Self-supervised video representation learning (167 papers)
- Video restoration with temporal priors (159 papers)
- Sign language and physiological video analysis (99 papers)
- Weakly supervised video anomaly detection (99 papers)
- Spatio-temporal predictive modeling (91 papers)
- Matching-based video object segmentation (86 papers)
- Learned optical flow estimation (73 papers)
- Learned video compression (68 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).