AI Paper Atlas

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

Bayesian Relational Memory for Semantic Visual Navigation

Yi Wu, Yuxin Wu, Aviv Tamar, Stuart Russell, Georgia Gkioxari, Yuandong Tian

ICCV 2019 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Vision-language navigation agents · 101 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 learned prior over room relations (vs uninformative prior) (ablation)BRM · House3D RoomNav H=300success rate / SPL: 20.9 / 39.4 -> 23.1 / 45.3Improved
add learned prior over room relations (vs uninformative prior) (ablation)BRM · House3D RoomNav H=1000success rate / SPL: 40.4 / 56.6 -> 41.1 / 57.5Improved
replace learned LSTM locomotion with oracle locomotion (ablation)BRM · House3D RoomNav H=1000success rate: 41.1 -> 88.6Improved
replace BRM planner with optimal planner (ablation)BRM with LSTM locomotion · House3D RoomNav H=1000success rate / SPL: 41.1 / 57.5 -> 46.3 / 62.5Improved
increase memory update interval N from 10 to 30 (ablation)BRM · House3D RoomNav H=1000success rate / SPL: 41.1 / 57.5 -> 29.7 / 35.2Hurt
increase memory update interval N from 10 to 50 (ablation)BRM · House3D RoomNav H=1000success rate / SPL: 41.1 / 57.5 -> 27.4 / 32.2Hurt
use ground-truth semantic signals instead of CNN detector (ablation)BRM · House3D RoomNav H=1000success rate / SPL: 41.1 / 57.5 -> 41.7 / 59.8No clear effect
add BRM planning on top of RNN controllerRNN controller · House3D RoomNav H=300success rate / SPL: 19.9 / 34.2 -> 23.1 / 45.3Improved
add BRM planning on top of RNN controllerRNN controller · House3D RoomNav H=1000success rate / SPL: 37.0 / 45.2 -> 41.1 / 57.5Improved
add BRM on top of RNN controller with terminate actionRNN controller · House3D RoomNav terminate action H=1000success rate / SPL: 14.5 / 16.3 -> 17.3 / 23.0Improved
BRM vs RNN control at planning distance 4, short horizonRNN controller · House3D RoomNav H=300, plan-steps 4success rate / SPL: 5.2 / 15.2 -> 5.8 / 18.6Mixed

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