AI Paper Atlas

Research map / Detection and segmentation / CNN features

Dual Transfer Learning for Event-Based End-Task Prediction via Pluggable Event to Image Translation

Lin Wang, Yujeong Chae, Kuk-Jin Yoon

ICCV 2021 · Detection and segmentation › CNN features · approach: Local features and context in CNNs · 48 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 the event-to-image translation (EIT) branch to CE training (ablation)DTL (DeepLabv3) · DDD17 (50ms)MIoU: 50.92 -> 53.87 (+2.95)Improved
add semantic consistency loss to CE + EIT (ablation)DTL (DeepLabv3) · DDD17 (50ms)MIoU: 53.87 -> 54.66Improved
add feature-level affinity transfer loss in the TL module (ablation)DTL (DeepLabv3) · DDD17 (50ms)MIoU: 50.92 -> 56.50 (+5.58)Improved
add prediction-level transfer loss on top of feature-level transfer (ablation)DTL (DeepLabv3) · DDD17 (50ms)MIoU: 56.50 -> 58.80Improved
use multi-channel event representation instead of voxel grid (ablation)DTL (DeepLabv3) · DDD17 (50ms)MIoU: 56.30 -> 58.80Improved
use multi-channel event representation instead of 6-channel (ablation)DTL (DeepLabv3) · DDD17 (50ms)MIoU: 57.68 -> 58.80Improved
add DTL framework to DeepLabv3 baselineDeepLabv3 · DDD17 (10ms window)MIoU: 41.61 -> 50.01 (+8.40)Improved
add DTL framework to DeepLabv3 baselineDeepLabv3 · DDD17 (250ms window)MIoU: 43.87 -> 52.96 (+9.09)Improved
add DTL framework to DeepLabv3 baselineDeepLabv3 · MVSEC segmentation (pseudo labels)MIoU: 50.53 -> 60.82 (+10.29)Improved
train DTL without ground truth labels using teacher supervisionDTL · DDD17 (50ms)MIoU: 56.52% MIoU, 6.60% above the baselineImproved
add DTL framework to depth baselineDepth baseline · MVSEC outdoor_day1Abs. Rel.: 0.33 -> 0.29Improved
add DTL framework to depth baselineDepth baseline · MVSEC outdoor_night1Abs. Rel.: 0.36 -> 0.30Improved

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