AI Paper Atlas

Research map / Detection and segmentation / CNN features

Event Camera Data Dense Pre-training

Yan Yang, Liyuan Pan, Liu liu

ECCV 2024 · Detection and segmentation › CNN features · approach: Spiking event-camera detection · 10 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 context-level similarity loss to pre-training (ablation)Event dense pre-training, Swin-T/7 on E-TartanAir · DSEC semantic segmentationmIoU / mACC (%): 55.556/63.486 -> 61.250/69.620Improved
add context-level loss to ECDP objective (ablation)ECDP (ViT-S/16) · DSEC semantic segmentationmIoU / mACC (%): 52.517/60.553 -> 53.826/61.008Improved
pre-train on synthetic E-TartanAir instead of N-ImageNet (ablation)Event dense pre-training (all backbones) · DSEC semantic segmentationmIoU (%): 56.654 -> 61.250 (Swin-T/7)Improved
increase number of contexts beyond 8 (ablation)Context-level similarity loss · DSEC semantic segmentationmIoU / mACCHurt
increase pre-training epochs beyond 300 (ablation)Event dense pre-training · DSEC semantic segmentationmIoU / mACCNo clear effect
use Swin-T/7 backbone instead of ViT-S/16Event dense pre-training · DSEC semantic segmentationmIoU (%): 55.729 -> 61.250Improved
event-only self-supervised pre-training vs supervised depth pre-trainingOurs vs HMNet · MVSEC depth (no cutoff)RMS: 7.534 -> 6.805Improved
pre-train with proposed method vs IDNet previous bestOurs (Swin-T/7) · DSEC-FlowEPE: 0.719 -> 0.697Improved

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