AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Diffusion-TTA: Test-time Adaptation of Discriminative Models via Generative Feedback

Mihir Prabhudesai, Tsung-Wei Ke, Alexander Cong Li, Deepak Pathak, Katerina Fragkiadaki

NeurIPS 2023 · Image generation › Fast diffusion sampling · approach: Diffusion models as perception tools · 5 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 a single random timestep and noise latent for diffusion loss TTA (ablation)Diffusion-TTA (ConvNext-Tiny) · ImageNet-Aaccuracy: 22.7 -> 19.8Hurt
add timestep augmentation with batch size 180 (ablation)Diffusion-TTA (ConvNext-Tiny) · ImageNet-Aaccuracy: 19.8 -> 24.5Improved
add random noise latents per timestep (ablation)Diffusion-TTA (ConvNext-Tiny) · ImageNet-Aaccuracy: 24.5 -> 24.7Improved
also adapt the diffusion model weights (ablation)Diffusion-TTA (ConvNext-Tiny) · ImageNet-Aaccuracy: 24.7 -> 25.8Improved
randomly initialize the classifier instead of pretrained weights (ablation)Diffusion-TTA (ResNet18) · ImageNetaccuracy: 68.4 -> 0.7Hurt
adapt only logits with diffusion objective, no classifier (ablation)Diffusion-TTA (ResNet18) · ImageNet-Raccuracy: 37.0 -> 31.0Hurt
adapt only BN layers (ablation)Diffusion-TTA (ResNet18) · ImageNetaccuracy: 68.4 -> 69.0No clear effect
adapt only last FC layer (ablation)Diffusion-TTA (ResNet18) · ImageNetaccuracy: 68.4 -> 68.6No clear effect
also adapt the diffusion model (DiT) in addition to the classifier (ablation)Diffusion-TTA (ResNet18) · ImageNetaccuracy: 73.0 -> 78.2Improved
apply Diffusion-TTA to CLIP-ViT-B/16CLIP-ViT-B/16 · CIFAR-100accuracy: 69.4 -> 69.0Hurt
apply Diffusion-TTA to CLIP-ViT-L/14CLIP-ViT-L/14 · Food101accuracy: 93.1 -> 93.1No clear effect
Diffusion-TTA vs TENT online on ConvNext-TinyConvNext-Tiny · ImageNet-Aaccuracy: 10.6 -> 25.8Improved

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