Research map / Video / Video segmentation and tracking
A Bit is All You Need! Efficient Video Capture via Single Bit Imaging
Kanchana Vaishnavi Gandikota, Michael Moeller, Andreas Kolb, Bhaskar Choubey, Paramanand Chandramouli
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 |
|---|---|---|---|
| permute uniform thresholds to be non-smooth instead of smoothly varying (ablation) | Aswin 1-bit video recovery · GoPro test | PSNR (dB): 30.32 -> 32.20 (no-overlap); 31.31 -> 32.75 (overlap) | Improved |
| use empirical-quantile thresholds instead of uniform non-smooth (ablation) | Aswin 1-bit video recovery · GoPro test | PSNR (dB): 32.20 -> 32.24 (no-overlap) | No clear effect |
| fit Beta distribution for thresholds instead of uniform non-smooth (ablation) | Aswin 1-bit video recovery · GoPro test | PSNR (dB): 32.20 -> 31.15 (no-overlap) | Hurt |
| learn thresholds jointly with network weights (ablation) | Aswin 1-bit video recovery · GoPro test | PSNR (dB): 32.20 -> 32.15 (no-overlap) | No clear effect |
| increase smooth threshold sequence length from 8 to 10 (ablation) | Aswin 1-bit video recovery · GoPro test | PSNR (dB): 29.23 -> 29.29 | No clear effect |
| increase smooth threshold sequence length from 10 to 20 distinct values (ablation) | Aswin 1-bit video recovery · GoPro test | PSNR (dB): 29.29 -> 26.50 | Hurt |
| repeat the 10-threshold sequence twice instead of using 10 or 20 distinct thresholds (ablation) | Aswin 1-bit video recovery · GoPro test | PSNR (dB): 29.29 -> 30.32 | Improved |
| scale binary input frames by the threshold value (ablation) | Aswin 1-bit video recovery · GoPro test | PSNR (dB): 31.03 -> 32.20 (no-overlap) | Improved |
| use symmetric scaling (0->tau/2, 1->(1+tau)/2) instead of 0->0, 1->tau (ablation) | Aswin 1-bit video recovery · GoPro test | PSNR (dB): 32.20 -> 32.17 (no-overlap) | No clear effect |
| add overlap strategy (replace boundary frames with shifted reconstruction) (ablation) | Aswin 1-bit video recovery · GoPro test | PSNR (dB): 32.24 -> 32.77 | Improved |
| fine-tune with Gaussian noise augmentation | EfficientSCI++ · GoPro test with noise | PSNR (dB): 31.27 (sigma=0) -> 33.02 (sigma=0.05) | Improved |
| test on hardware-generated frames with non-ideal comparator | EfficientSCI++ fine-tuned on noise · GoPro test | PSNR (dB): 31.27 -> 30.77 | Hurt |
Closest papers by meaning
- A Little Bit More: Bitplane-Wise Bit-Depth Recovery (TPAMI 2021 · similarity 0.86)
- One-Bit Time-Resolved Imaging (TPAMI 2020 · similarity 0.86)
- 3One2: One-Step Regression plus One-Step Diffusion for One-Hot Modulation in Dual-Path Video Snapshot Compressive Imaging (AAAI 2026 · similarity 0.85)
- bit2bit: 1-bit quanta video reconstruction via self-supervised photon prediction (NeurIPS 2024 · similarity 0.84)
- Towards HDR and HFR Video from Rolling-Mixed-Bit Spikings (CVPR 2024 · similarity 0.83)
- Video Depth Anything: Consistent Depth Estimation for Super-Long Videos (CVPR 2025 · similarity 0.83)
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.
