AI Paper Atlas

Research map / Federated learning and privacy / Differential privacy

Unsafe2Safe: Controllable Image Anonymization for Downstream Utility

Minh Dinh, SouYoung Jin

CVPR 2026 · Federated learning and privacy › Differential privacy · approach: Private training with public data · 1 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
fine-tune InstructPix2Pix on the anonymization dataset (c_edit prior) (ablation)U2S InstructPix2Pix · Caltech101 accuracytop-1 accuracy: 94.315 -> 95.116Improved
fine-tune InstructPix2Pix on the anonymization dataset (c_edit prior) (ablation)U2S InstructPix2Pix · MIT Indoor67FaceSim: 0.2909 -> 0.2735Improved
add Safe Attention to the fine-tuned InstructPix2Pix (ablation)U2S InstructPix2Pix fine-tuned · MIT Indoor67FaceSim: 0.2735 -> 0.2465Improved
add Safe Attention to the fine-tuned InstructPix2Pix (ablation)U2S InstructPix2Pix fine-tuned · Caltech101race entropy: 0.7997 -> 0.8306Improved
use the edit instruction c_edit instead of public caption as FlowEdit prior (ablation)U2S FlowEdit · MIT Indoor67top-1 accuracy: 80.746 -> 77.090Hurt
use LLM-composed prior instead of c_public with FlowEdit (ablation)U2S FlowEdit · Caltech101FaceSim: 0.4881 -> 0.3428Improved
use LLM-composed prior instead of class prior with FlowEdit (ablation)U2S FlowEdit · Caltech101race entropy: 0.4051 -> 0.8751Improved
use LLM-composed prior instead of c_edit with FlowEdit (ablation)U2S FlowEdit · Caltech101top-1 accuracy: 94.792 -> 92.884Hurt
train on only the safe subset (discard flagged images) instead of anonymizing them (ablation)ImageMAE classifier · Caltech101top-1 accuracy: 94.27 -> 83.49Hurt
train on only the safe subset (discard flagged images) instead of anonymizing them (ablation)ImageMAE classifier · MIT Indoor67top-1 accuracy: 83.88 -> 51.12Hurt
train VQA model on U2S (FE, LLM) anonymized images instead of rawQwen3VL-2B on OK-VQA · OK-VQAVQA accuracy: 0.6064 -> 0.7093Improved
train captioner on U2S (FE, LLM) anonymized images instead of rawBLIP-2 on MS-COCO · MS-COCO captioningCIDEr: 1.413 -> 1.363No clear effect

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