AI Paper Atlas

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

THDA: Treasure Hunt Data Augmentation for Semantic Navigation

Oleksandr Maksymets, Vincent Cartillier, Aaron Gokaslan, Erik Wijmans, Wojciech Galuba, Stefan Lee, Dhruv Batra

ICCV 2021 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Vision-language navigation agents · 74 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 RGB-D input with depth plus semantic segmentation and goal segmentationRGB-D ObjectNav agent · ObjectNav valSuccess (%): 8.0 -> 10.2Improved
add exploration reward to sparse success reward (ablation)Sparse Success · ObjectNav valSuccess (%): 0.0 -> 10.8Improved
use ExploreTillSeen reward instead of dense path-efficiency reward (ablation)Depth +SemSeg +GoalSeg · ObjectNav valSuccess (%): 10.2 -> 14.5Improved
use ExploreTillSeen reward instead of dense path-efficiency reward (ablation)Depth +SemSeg +GoalSeg · ObjectNav valSPL (%): 3.3 -> 4.8Improved
use sparse success-only reward instead of dense reward (ablation)Depth +SemSeg +GoalSeg · ObjectNav valSuccess (%): 10.2 -> 0.0Hurt
remove SemSeg, keep depth and GoalSeg (ablation)ExploreTillSeen agent · ObjectNav valSuccess (%): 14.5 -> 20.0Improved
remove SemSeg, keep depth and GoalSeg (ablation)ExploreTillSeen agent · ObjectNav valSPL (%): 4.8 -> 6.5Improved
replace predicted goal segmentation with ground-truth segmentationExploreTillSeen Depth + GoalSeg · ObjectNav valSuccess (%): 20.0 -> 40.7Improved
add THDA pretraining then fine-tune on ObjectNav (ablation)Depth + GoalSeg · ObjectNav valSuccess (%): 20.0 -> 28.4Improved
add THDA pretraining then fine-tune on ObjectNav (ablation)Depth + GoalSeg · ObjectNav valSPL (%): 6.5 -> 11.0Improved
THDA pretraining only, zero-shot without ObjectNav trainingDepth + GoalSeg · ObjectNav valSuccess (%): 11.4Mixed
Depth + GoalSeg + THDA vs previous best Habitat leaderboard entrySemExp / SRCB-robot-sudoer · ObjectNav test-standardSuccess (%): 17.9 -> 27.4Improved

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