AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Norm-guided latent space exploration for text-to-image generation

Dvir Samuel, Rami Ben-Ari, Nir Darshan, Haggai Maron, Gal Chechik

NeurIPS 2023 · Image generation › Text-to-image diffusion · approach: Multi-concept composition · 2 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 norm-prior-guided interpolation instead of SLERPStable Diffusion seed interpolation · ImageNet1kAcc: 30.41 -> 51.59Improved
Use norm-prior-guided interpolation instead of SLERPStable Diffusion seed interpolation · ImageNet1kFID: 18.44 -> 6.78Improved
Use a norm-prior-induced centroid instead of sphere projectionStable Diffusion seed centroid · ImageNet1kAcc: 40.81 -> 67.24Improved
Use a norm-prior-induced centroid instead of sphere projectionStable Diffusion seed centroid · ImageNet1kFID: 14.28 -> 5.48Improved
Initialize SeedSelect with NAO-centroid seedsSeedSelect · ImageNet1k in LAION2B, totalTotal Acc: 91.3 -> 94.3Improved
Initialize SeedSelect with NAO-centroid seedsSeedSelect · ImageNet1k in LAION2B, totaloptimization time (sec): 298 -> 29Improved
Use NAO-centroid augmentation instead of SeedSelectFew-shot classifier · miniImageNet, 5-way 5-shotaccuracy: 92.08±0.7 -> 93.21±0.6Improved
Use NAO-centroid augmentation instead of SeedSelectFew-shot classifier · CIFAR-FS, 5-way 5-shotaccuracy: 94.87±0.4 -> 94.85±0.5No clear effect
Use NAO-centroid augmentation instead of SeedSelectFew-shot classifier · CUB, 5-way 5-shotaccuracy: 96.01±0.4 -> 97.92±0.2Improved
Use NAO-centroid augmentation instead of SeedSelectLong-tail classifier · iNaturalistaccuracy: 74.8±0.7 -> 75.3±0.7No clear effect
Use NAO-centroid augmentation instead of SeedSelectLong-tail classifier · ImageNet-LTaccuracy: 74.9±0.5 -> 78.9±0.1Improved
Initialize SeedSelect with NAO seeds rather than random seedsSeedSelect · few-shot recognitionaccuracyImproved

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