AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Pre-Training Without Natural Images

Hirokatsu Kataoka, Kazushige Okayasu, Asato Matsumoto, Eisuke Yamagata, Ryosuke Yamada, Nakamasa Inoue, Akio Nakamura, Yutaka Satoh

IJCV 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 74 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
render fractals with 3x3 patches instead of 1x1 points (ablation)FractalDB · CIFAR-10accuracy (%): 87.4 -> 92.1Improved
render fractals with 3x3 patches instead of 1x1 points (ablation)FractalDB · CIFAR-100accuracy (%): 66.1 -> 72.0Improved
use fixed patch instead of random patch pattern (ablation)FractalDB · CIFAR-100accuracy (%): 72.0 -> 73.6Improved
filling rate 0.05 to 0.15 (ablation)FractalDB · CIFAR-10accuracy (%): 91.8, 92.0, 91.7No clear effect
raise filling rate to 0.25 (ablation)FractalDB · CIFAR-100accuracy (%): 72.3 -> 63.2Hurt
widen intra-category weight interval from 0.1 to 0.4 (ablation)FractalDB · CIFAR-10accuracy (%): 92.1 -> 92.7Improved
widen intra-category weight interval to 0.5 (ablation)FractalDB · CIFAR-10accuracy (%): 92.7 -> 91.8Hurt
increase image size from 256 to 1024 (ablation)FractalDB · CIFAR-100accuracy (%): 73.6 -> 68.6Hurt
increase number of dots from 100k to 400k (ablation)FractalDB · CIFAR-10accuracy (%): 91.3 -> 90.4Hurt
assign categories by DeepCluster k-means instead of IFS parameters (ablation)FractalDB · CIFAR-100accuracy (%): 77.3 vs 57.0Hurt
freeze conv1 during fine-tuning (ablation)FractalDB-1k · CIFAR-100accuracy (%): 75.7 -> 72.2Hurt
freeze conv1-5 during fine-tuning (ablation)FractalDB-1k · CIFAR-10accuracy (%): 93.4 -> 49.4Hurt

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