Research map / Architectures and efficiency / CNN design
Seeing Into Darkness: Scotopic Visual Recognition
Bo Chen, Pietro Perona
Check 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 time-adapted first-layer features instead of time-normalized rates (ablation) | WaldNet versus Rate classifier · MNIST and CIFAR10 | speed-accuracy tradeoff | Improved |
| Train with scotopic images and use an appropriate stopping criterion instead of retrofitting a photopic classifier (ablation) | WaldNet versus Photopic classifier · MNIST and CIFAR10 | speed-accuracy tradeoff | Improved |
| Replace a fixed interrogation duration with free-response stopping (ablation) | Scotopic classifiers · MNIST and CIFAR10 | median photon counts: 3× reduction | Improved |
| Learn dynamic stopping thresholds instead of using constant thresholds (ablation) | Scotopic classifiers · MNIST and CIFAR10 | average PPP at a given error rate | Improved |
| Use learned dynamic thresholds instead of constant thresholds (ablation) | Scotopic classifiers · CIFAR10 | median PPP | Hurt |
| Use a single time-adapted network instead of an ensemble of specialists (ablation) | WaldNet versus Ensemble · MNIST, free-response | speed-accuracy tradeoff | No clear effect |
| Use a single time-adapted network instead of an ensemble of specialists (ablation) | WaldNet versus Ensemble · CIFAR10, lowlight conditions (≤ 22 PPP) | speed-accuracy tradeoff | Improved |
| Increase dark current to 11% (ablation) | WaldNet · MNIST and CIFAR10 | exposure time: doubles | Hurt |
| Increase read noise to 22% (ablation) | WaldNet · MNIST | accuracy | No clear effect |
| Increase read noise to 22% (ablation) | WaldNet · CIFAR10 | accuracy | Hurt |
| Increase read noise to 50% (ablation) | WaldNet · MNIST and CIFAR10 | accuracy | Hurt |
| Increase rotational jitter to 22° at 220 PPP (ablation) | WaldNet · MNIST and CIFAR10 | accuracy | Hurt |
Closest papers by meaning
- Quanta Neural Networks: From Photons to Perception (ICCV 2025 · similarity 0.86)
- Reconstructing Transient Images From Single-Photon Sensors (CVPR 2017 · similarity 0.85)
- Finding Visual Saliency in Continuous Spike Stream (AAAI 2024 · similarity 0.82)
- Brain-Inspired Spiking Neural Networks for Energy-Efficient Object Detection (CVPR 2025 · similarity 0.82)
- Optical Flow for Spike Camera with Hierarchical Spatial-Temporal Spike Fusion (AAAI 2024 · similarity 0.82)
- SFOD: Spiking Fusion Object Detector (CVPR 2024 · similarity 0.82)
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.
