AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Semantic Jitter: Dense Supervision for Visual Comparisons via Synthetic Images

Aron Yu, Kristen Grauman

ICCV 2017 · Image generation › Autoregressive and latent generation · approach: Classical generative image synthesis · 136 citations (OpenAlex)

Read the paper

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 human-verified semantic jitter pairs (DSynth) to real training pairs (ablation)DeepSTN · PFSmileaccuracy: 81.52 -> 85.78Improved
add automatically labeled semantic jitter pairs (DSynth-Auto) (ablation)RankSVM · PFSmileaccuracy: 69.29 -> 75.00Improved
add automatically labeled semantic jitter pairs (DSynth-Auto) (ablation)DeepSTN · PFSmileaccuracy: 81.52 -> 84.36Improved
use semantic jitter instead of low-level jitter (ablation)DeepSTN · PFSmileaccuracy: 80.09 -> 85.78Improved
add low-level jitter to real pairs (ablation)DeepSTN · PFSmileaccuracy: 81.52 -> 80.09No clear effect
add pseudo-pairs from generator attribute strengths (Real+) (ablation)DeepSTN · PFSmileaccuracy: 81.52 -> 80.84No clear effect
add pseudo-pairs from generator attribute strengths (Real+) (ablation)RankSVM · PFSmileaccuracy: 69.29 -> 68.95No clear effect
replace human-verified (DSynth) with auto-labeled (DSynth-Auto) synthetic pairs (ablation)DeepSTN · UT-Zap50K lexicon, Comfortaccuracy: 82.18 -> 87.27Improved
add DSynth-Auto synthetic pairs (ablation)DeepSTN · UT-Zap50K lexicon, Comfortaccuracy: 84.95 -> 87.27Improved
add low-level jitter to real pairs (ablation)DeepSTN · UT-Zap50K lexicon, Comfortaccuracy: 84.95 -> 81.94Hurt
add human-verified synthetic pairs (DSynth) (ablation)DeepSTN · Synthetic test pairs, Sleekaccuracy: 78.76 -> 98.12Improved
replace DeepSTN baseline with DeepSTN trained with DSynth-AutoDeepSTN · UT-Zap50K-2 (Open)accuracy: 70.73 -> 72.18Improved

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