AI Paper Atlas

Research map / 3D vision / Depth and stereo

Photon-Starved Scene Inference Using Single Photon Cameras

Bhavya Goyal, Mohit Gupta

ICCV 2021 · 3D vision › Depth and stereo · approach: Spike and event camera reconstruction · 14 citations (OpenAlex)

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 changedWhereMeasuredEffect
add photon scale space images to training (ablation)Vanilla Net (ResNet-18) · CUB-200-2011 S1top-1 accuracy: 21.35 -> 28.92Improved
add feature consistency (MSE) loss across PPP levels (ablation)Vanilla Net w/ Photon Scaled Images · CUB-200-2011 S1top-1 accuracy: 28.92 -> 42.37Improved
use Photon Net instead of Curriculum Learning (ablation)CUB-200-2011 classifier · CUB-200-2011 S1top-1 accuracy: 33.72 -> 42.37Improved
use Photon Net instead of N-step Student Teacher Learning (ablation)CUB-200-2011 classifier · CUB-200-2011 S1top-1 accuracy: 35.79 -> 42.37Improved
BM3D denoising of input images (ablation)Vanilla Net · CUB-200-2011 S1top-1 accuracy: 21.35 -> 25.52Improved
use Photon Net instead of Joint Denoisingclassifier · CUB-200-2011 S1top-1 accuracy: 27.21 -> 42.37Improved
use Photon Net instead of Joint Denoisingclassifier · CARS-196 S1top-1 accuracy: 34.51 -> 64.23Improved
use Photon Net instead of Student-Teacher Learningclassifier · CARS-196 S10top-1 accuracy: 73.51 -> 78.97Improved
use Photon Net instead of Joint DenoisingDenseDepth · NYUv2 S1rel: 0.209 -> 0.183Improved
use Photon Net instead of Joint DenoisingDenseDepth · NYUv2 S10delta1: 0.763 -> 0.793Improved
increase number of photon scale space levels from 2 to 5 (ablation)Photon Net · CUB-200-2011 and CARS-196 S1top-1 accuracyImproved
set MSE loss weight lambda to 25.0 (vs 0 to 50.0 range) (ablation)Photon Net · CUB-200-2011 S1top-1 accuracyImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy