AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

Visual Graph Memory With Unsupervised Representation for Visual Navigation

Obin Kwon, Nuri Kim, Yunho Choi, Hwiyeon Yoo, Jeongho Park, Songhwai Oh

ICCV 2021 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Vision-language navigation agents · 82 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
replace stacked memory (SMT) with graph memory VGMSMT · Gibson image-goal navigation (Overall)success rate: 0.68 -> 0.76Improved
replace stacked memory (SMT) with graph memory VGMSMT · Gibson image-goal navigation (Overall)SPL: 0.56 -> 0.64Improved
replace metric map (Exp4nav) with VGMExp4nav · Gibson image-goal navigation (Overall)success rate: 0.59 -> 0.76Improved
replace metric map (Exp4nav) with VGMExp4nav · Gibson image-goal navigation (Overall)SPL: 0.51 -> 0.64Improved
use VGM instead of SMT stored observationsSMT · Gibson image-goal navigationmemory elements: 96% fewer memory elementsImproved
lower similarity threshold s_th from 0.75 to 0.30 (ablation)VGM · Gibson hard episodessuccess rate: 0.61 -> 0.60No clear effect
lower similarity threshold s_th from 0.75 to 0.30 (ablation)VGM · Gibson hard episodesSPL: 0.46 -> 0.41Hurt
lower similarity threshold s_th from 0.75 to 0.30 (ablation)VGM · Gibson hard episodeslocalization accuracy: 0.94 -> 0.97No clear effect
add pose sensor noise to metric-map and pose-based baselinesANS and Neural Planner · Gibson hard episodessuccess rateHurt
add pose sensor noise to VGMVGM · Gibson hard episodessuccess rateNo clear effect
add auxiliary losses (ablation)VGM · Gibsonsuccess rateImproved
train F_loc with supervised learning instead of unsupervised (ablation)VGM · Gibsonsuccess rate: about 8% improvementImproved

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