AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Visual Atoms: Pre-Training Vision Transformers With Sinusoidal Waves

Sora Takashima, Ryo Hayamizu, Nakamasa Inoue, Hirokatsu Kataoka, Rio Yokota

CVPR 2023 · Architectures and efficiency › Pruning and efficient tuning · approach: Efficient ViT adaptation · 25 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
Pre-train with VisualAtom-1k instead of RCDB-1kViT-T · CIFAR100accuracy: 82.2 -> 84.9Improved
Pre-train with VisualAtom-21k instead of RCDB-21kViT-B · CIFAR100accuracy: 82.9 -> 86.7Improved
Scale pre-training from VisualAtom-1k to VisualAtom-21kViT-B · Average of seven evaluation datasetsaccuracy: 89.1 -> 89.7Improved
Pre-train with VisualAtom-1k instead of ImageNet-21kViT-T · ImageNet-1k fine-tuning, 224²top-1 accuracy: 74.1 -> 74.2Improved
Use VisualAtom-21k rather than JFT-300M for pre-trainingViT-B · ImageNet-1k fine-tuning, 384²top-1 accuracy: 84.2 -> 83.7Hurt
Increase the maximum frequency beyond 20 (ablation)VisualAtom-1k · CIFAR100accuracy: 84.9 -> 84.3; 84.9 -> 84.1Hurt
Narrow the frequency range by raising its minimum (ablation)VisualAtom-1k · CIFAR100accuracy: 84.9 -> 84.8; 84.9 -> 84.4; 84.9 -> 83.6Hurt
Increase variation in orbit count from fixed/high counts to a broad range (ablation)VisualAtom-1k · CIFAR100accuracy: 84.3 -> 84.9Improved
Use low quantization values (3–200) rather than the baseline range (ablation)VisualAtom-1k · CIFAR100accuracy: 84.9 -> 84.6Hurt
Vary amplitude rather than fixing it at 0.5 (ablation)VisualAtom-1k · CIFAR100accuracy: 84.9 -> 84.9No clear effect
Increase fixed amplitude from 0.5 to 1.0 (ablation)VisualAtom-1k · CIFAR100accuracy: 84.9 -> 84.4Hurt

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