AI Paper Atlas

Research map / 3D vision / 3D generation

Learning Physical Intuition of Block Towers by Example

ICML 2016 · 3D vision › 3D generation · approach: Learning physics from video · 75 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's title only (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
Jointly predict block masks and fall outcome rather than train fall classification alonePhysNet on synthetic block towers · Synthetic block towersfall accuracy (%): 89.1 vs 86.5Helped
Use PhysNet rather than DeepMask for future mask predictionSynthetic block towers · Synthetic block towersMask IoU (%): 42.4 -> 75.4Helped
Use PhysNet rather than DeepMask for future mask predictionSynthetic block towers · Synthetic block towersLog likelihood/px: -0.299 -> -0.107Helped
Remove ImageNet pretraining from Googlenet (ablation)Googlenet · Real block towersfall accuracy (%): 69.0 ± 2.1 -> 59.2 ± 2.2Hurt
Remove ImageNet pretraining from Googlenet (ablation)Googlenet · Synthetic block towersfall accuracy (%): 86.5 -> 86.5No clear effect
Train without 4-block towers (ablation)Googlenet · Synthetic 4-block towersfall accuracy (%): 82.3 -> 71.3Hurt
Train without 3-block towers (ablation)PhysNet · Synthetic 3-block towersfall accuracy (%): 87.9 -> 84.5Hurt

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