AI Paper Atlas

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

Object-Oriented Dynamics Learning through Multi-Level Abstraction

Guangxiang Zhu, Jianhao Wang, Zhizhou Ren, Zichuan Lin, Chongjie Zhang

AAAI 2020 · Detection and segmentation › Open-set and few-shot detection · approach: Object-centric learning theory and video · 9 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
use the three-level MAOP framework for object-oriented dynamics learningaction-conditioned dynamics prediction · Monster Kong, 1-5, 3k samples, unseen environments0-error accuracy: 0.14 -> 0.94Improved
use MAOP instead of CDNAaction-conditioned dynamics prediction · Flappy Bird, 100 samples, Agent0-acc: 0.13 -> 0.83Improved
use MAOP instead of CDNAaction-conditioned dynamics prediction · Freeway, 100 samples0-acc: 0.42 -> 0.78Improved
use MAOP instead of CDNAaction-conditioned dynamics prediction · Krull, 3k samples0acc: 0.27 -> 0.14Hurt
use MAOP instead of OODPaction-conditioned dynamics prediction · Breakout, 3k samples0acc: 0.73 -> 0.52Hurt
use MCTS with the learned MAOP dynamics model instead of DQNplanning in unseen Monster Kong environments · Monster KongReward: 13.67 -> 38.19Improved
use the learned MAOP model rather than the real simulator for MCTSplanning in unseen Monster Kong environments · Monster KongReward: 38.41 -> 38.19No clear effect
replace DIS instance proposals with Gaussian-noise proposals (ablation)MAOP dynamics learning · Monster Kong, unseen environments, 3-to-5 generalization, Agent0-acc: 0.99 -> 0.60Hurt
replace DIS instance proposals with Poisson-noise proposals (ablation)MAOP dynamics learning · Monster Kong, unseen environments, 3-to-5 generalization, Agent0-acc: 0.99 -> 0.93Hurt
use 3 rather than 5 object masks (ablation)MAOP · Monster Kong, unseen environments, Agent0-error accuracy: 0.94 -> 0.90Hurt
use 8 rather than 5 object masks (ablation)MAOP · Monster Kong, unseen environments, Agent0-error accuracy: 0.94 -> 0.91Hurt
train with 3k rather than 300k samplesMAOP versus CDNA · Monster Kong, unseen environments, Agent0-error accuracy: CDNA: 300k samples, 0.26; MAOP: 3k samples, 0.94Improved

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