AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

Unsupervised Object-Based Transition Models For 3D Partially Observable Environments

Antonia Creswell, Rishabh Kabra, Christopher P. Burgess, Murray Shanahan

NeurIPS 2021 · Detection and segmentation › Open-set and few-shot detection · approach: Object-centric learning theory and video · 0 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
remove AlignNet (OT vs OAT) (ablation)OAT · 3D RoomUnroll ARI: 0.42 -> 0.12Hurt
remove AlignNet (ablation)OAT · 3D RoomEncoding ARI: 0.62 -> 0.59Hurt
remove AlignNet (ablation)OAT · 3D RoomUnroll Pixel Error: 0.0121 -> 0.0143Hurt
OAT versus OP3 baselineOAT · 3D RoomUnroll ARI: 0.33 -> 0.42Improved
OAT versus OP3 baselineOAT · 3D RoomEncoding ARI: 0.32 -> 0.62Improved
OAT versus OP3 baselineOAT · 3D RoomUnroll Pixel Error: 0.0132 -> 0.0121Improved
use pixel-level loss instead of object-level loss with aligned inputs (ablation)slot-wise transition model · Ground-truth mask ablationobject-level error: 0.0982 -> 1.04Hurt
use unaligned inputs with object-level loss (ablation)slot-wise transition model · Ground-truth mask ablationobject-level error: 0.0982 -> 19.9Hurt
use unaligned inputs with pixel-level loss (ablation)slot-wise transition model · Ground-truth mask ablationobject-level error: 1.04 -> 229Hurt
transformer followed by slot-wise LSTM versus Transformer XL, LSTM, Deep LSTM, recurrent transformer (ablation)transition core · Ground-truth mask ablationobject and pixel errorImproved
lower transition loss weight zeta from 10 to 1.0 (ablation)OAT · 3D Roomperformance: 10 -> 1.0No clear effect
mask KL weight gamma 0.5 instead of 0.05 (ablation)OAT MONet · 3D Roomqualitative quality: 0.05 -> 0.5No clear effect

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