Research map / 3D vision / Depth and stereo
Spike and event camera reconstruction
Reconstructs high-speed scenes from spike, single-photon and neuromorphic sensor data.
50 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Depth and stereo, keyword cluster “light · imaging · illumination”
State History Tried Baselines 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.
10 papers in 2023–24 and 13 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 3.3% to 3.2%.
Open questions the data points to
- Papers report HDR-CNN on Test set (PSNR-γ) anywhere from 34.06 to 42.9: comparisons against it depend on whose number you take.
- Papers report LDR on Test set (PSNR-γ) anywhere from 33.57 to 38.9: comparisons against it depend on whose number you take.
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 · The Next Best Underwater View (CVPR 2016 · 50 citations)
- 2017 · Reconstructing Transient Images From Single-Photon Sensors (CVPR 2017 · 111 citations)
- 2019 · Reflection Removal Using a Dual-Pixel Sensor (CVPR 2019 · 81 citations)
- 2020 · Deep Optics for Single-Shot High-Dynamic-Range Imaging (CVPR 2020 · 152 citations)
- 2021 · Spk2ImgNet: Learning To Reconstruct Dynamic Scene From Continuous Spike Stream (CVPR 2021 · 75 citations)
- 2022 · Raw High-Definition Radar for Multi-Task Learning (CVPR 2022 · 120 citations)
- 2023 · Learning to Super-resolve Dynamic Scenes for Neuromorphic Spike Camera (AAAI 2023 · 18 citations)
- 2024 · Boosting Spike Camera Image Reconstruction from a Perspective of Dealing with Spike Fluctuations (CVPR 2024 · 16 citations)
- 2025 · All-Day Multi-Camera Multi-Target Tracking (CVPR 2025 · 8 citations)
- 2026 · Kaleidoscopic Scintillation Event Imaging (CVPR 2026 · 0 citations)
Earliest papers here
- The Next Best Underwater View (CVPR 2016 · 50 citations)
- Reconstructing Transient Images From Single-Photon Sensors (CVPR 2017 · 111 citations)
- Computational Imaging on the Electric Grid (CVPR 2017 · 41 citations)
What has been tried
41 papers here report 339 measured changes (a component added, removed or swapped, compared with a baseline): helped 78%, no clear effect 1%, hurt 17%, mixed 4%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| replace binary spike input with multi-order DSFT fusion | REDS-SCIR · PSNR | 38.44 -> 39.76 | Helped | Boosting Spike Camera Image Reconstruction from a Perspective of Dealing with Spike Fluctuations (CVPR 2024 · 16 citations) |
| add the multi-scale 3D encoder in CSR | DSPK, GN σ = 15 · PSNR | 30.99dB -> 32.01dB | Helped | Joint Demosaicing and Denoising for Spike Camera (AAAI 2024 · 13 citations) |
| Spatiotemporal chunk change detection (Sec 4.2) instead of other generalized designs | Rate-distortion analysis · PSNR | Helped | Generalized Event Cameras (CVPR 2024 · 10 citations) | |
| add BSSR | REDS-based dataset ×2 · PSNR | 33.04 -> 33.39 | Helped | Super-Resolution Reconstruction from Bayer-Pattern Spike Streams (CVPR 2024 · 9 citations) |
| add the ADMF fusion channels to simple averaged RGB-infrared fusion | M3Track · MOTA | 58.98 -> 68.23 | Helped | All-Day Multi-Camera Multi-Target Tracking (CVPR 2025 · 8 citations) |
| remove joint motion estimation | RBSS · PSNR | 31.42, 29.65, 28.08 -> 31.27, 29.50, 27.91 | Hurt | Self-Supervised Learning for Color Spike Camera Reconstruction (CVPR 2025 · 3 citations) |
| Use the two-stage RAW pipeline rather than one-stage RAW reconstruction | UHDR dataset ×50, ×100, and ×200 · PSNR | 25.137 -> 27.598; 24.896 -> 27.327; 24.632 -> 27.091 | Helped | UltraLED: Learning to See Everything in Ultra-High Dynamic Range Scenes (NeurIPS 2025 · 2 citations) |
| supervise reconstruction with both color and structure components rather than the color component alone | Kalantari et al. test · PSNR-u | 34.45 -> 43.68 | Helped | Self-Supervised High Dynamic Range Imaging with Multi-Exposure Images in Dynamic Scenes (ICLR 2024 · 2 citations) |
| Use camera-imaging-aware L2H-CC rather than an MLP | Synthetic datasets · PSNR (HDR) | 19.02 -> 38.57 | Helped | High Dynamic Range Novel View Synthesis with Single Exposure (ICML 2025 · 0 citations) |
| Add Cross-Modal Fusion (CF) | Simulated event data · SSIM | 0.603 -> 0.604 | No clear effect | EvFocus: Learning to Reconstruct Sharp Images from Out-of-Focus Event Streams (ICML 2025) |
Baseline numbers that disagree
The same baseline on the same benchmark and metric, reported more than 5% apart by different papers here. Training budgets, data splits and reimplementations differ, so state which number you compare with.
| Baseline | Benchmark | Metric | Reported | Papers |
|---|---|---|---|---|
| HDR-CNN | Test set | PSNR-γ | 34.06 – 42.9 | 2 papers, e.g. CVPR 2020 and CVPR 2020 |
| LDR | Test set | PSNR-γ | 33.57 – 38.9 | 2 papers, e.g. CVPR 2020 and CVPR 2020 |
| HDR-CNN | Test set | HDR-VDP-2 | 54.34 – 58.6 | 2 papers, e.g. CVPR 2020 and CVPR 2020 |
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 23 papers; from OpenAlex and the papers' affiliations).
- Peking University · 7 papers
- Harbin Institute of Technology · 4 papers
- University of Electronic Science and Technology of China · 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
- Neural light field and illumination imaging (107 papers)
- Non-line-of-sight imaging (93 papers)
- Snapshot hyperspectral imaging (38 papers)
Elsewhere in the topic
- Feed-forward 3D reconstruction (133 papers)
- Self-supervised monocular depth (116 papers)
- Scene coordinate regression for localization (112 papers)
- Learned optical and scene flow (108 papers)
- Cost-volume multi-view stereo (107 papers)
- Camera model and calibration for SfM (99 papers)
- Dense depth prediction and completion (85 papers)
- Foundation models for monocular geometry (81 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).