AI Paper Atlas

Research map / Detection and segmentation / CNN features

Learning to Exploit Multiple Vision Modalities by Using Grafted Networks

Yuhuang Hu, Tobi Delbruck, Shih-Chii Liu

ECCV 2020 · Detection and segmentation › CNN features · approach: Region-based and query-based detectors · 64 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
graft a trained thermal front end instead of feeding thermal frames into the pretrained networkYOLOv3 · FLIR thermal drivingAP50: 39.92 -> 45.27Improved
use thermal grafted network instead of intensity framesYOLOv3 · FLIR thermal drivingAP50: 30.36 -> 45.27Improved
supervised training of 47M parameters instead of grafting 3.2M parametersFaster R-CNN · FLIR thermal drivingAP50: 45.27 -> 53.97Improved
use DVS-10 instead of DVS-3 event slicesYOLOv3-GN · MVSEC car detectionAP50: 70.14 -> 70.35Improved
train on only 40% of the training samples (ablation)YOLOv3-GN · MVSEC car detectionAP50: 70.35 -> 66.75Hurt
train on only 10% of the training samples (ablation)YOLOv3-GN · MVSEC car detectionAP50: 70.35 -> 47.88Hurt
feed event volume to the pretrained intensity network without a grafted front end (ablation)YOLOv3-N · MVSEC car detectionAP50: 73.53 -> almost 0Hurt
combine event-driven and intensity predictions with NMSYOLOv3 · MVSEC car detectionAP50: 73.53 -> 75.45Improved
use deeper front end S3 instead of shallower S1 or S2 (ablation)YOLOv3-GN · FLIR thermal and MVSECAP50Improved
use middle net S5 instead of S4 (ablation)YOLOv3-GN · FLIR thermal and MVSECAP50No clear effect
add the style loss FSL to FRL and FEL (ablation)YOLOv3-GN · FLIR thermal and MVSECAP50No clear effect
train front end self-supervised instead of supervised on labeled N-MNIST (ablation)LeNet · N-MNISTAP50: 1.11 -> 1.47Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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