Research map / Image generation / Editing and style transfer
ID-Sim: An Identity-Focused Similarity Metric
Julia Chae, Nicholas Kolkin, Jui-Hsien Wang, Richard Zhang, Sara Beery, Cusuh Ham
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add patch-level supervision (ablation) | ID-Sim · PerSAM on PODS | mAP: 0.214 -> 0.436 | Improved |
| add patch-level supervision (ablation) | ID-Sim · DeepFashion2 (patch) | mAP: 0.4765 -> 0.7967 | Improved |
| add patch-level supervision (ablation) | ID-Sim · PODS (patch) | mAP: 0.8181 -> 0.7907 | Hurt |
| finetune DINOv3 into ID-Sim for PerSAM | PerSAM · PODS | mAP: 0.153 -> 0.436 | Improved |
| balance training data (ablation) | ID-Sim training set · validation score | validation score: 0.693 -> 0.752 | Improved |
| filter noisy instance-level data (ablation) | ID-Sim training set · validation score | validation score: 0.752 -> 0.890 | Improved |
| add positive edit synthetic data (ablation) | ID-Sim training set · validation score | validation score: 0.890 -> 0.937 | Improved |
| add identity-altering negative edits (ablation) | ID-Sim training set · validation score | validation score: 0.937 -> 0.965 | Improved |
| scale training set from 5k to 20k instances (ablation) | ID-Sim · Real Instance Validation | accuracy: 0.860 -> 0.880 | No clear effect |
| scale training set to 30k instances (ablation) | ID-Sim · Identity Validation | ROC AUC: 0.98 -> 0.91 | Hurt |
| make subset 1 dominant instead of even mix (ablation) | ID-Sim · validation sets | accuracy / ROC AUC: 0.8715 -> 0.8905 acc, 0.97 -> 0.95 AUC | Mixed |
Closest papers by meaning
- MaSC: A Masked Similarity Metric for Evaluating Concept-Driven Generation (NeurIPS 2026 · similarity 0.88)
- Illuminating Visual Identity in Universal Multimodal Embeddings (CVPR 2026 · similarity 0.87)
- Omni-ID: Holistic Identity Representation Designed for Generative Tasks (CVPR 2025 · similarity 0.86)
- ARGUS: Stacked Multi-View Identity Mosaic Injection for Subject-Preserving Video Generation (NeurIPS 2026 · similarity 0.86)
- ID-Patch: Robust ID Association for Group Photo Personalization (CVPR 2025 · similarity 0.86)
- FaceA-Net: Facial Attribute-Driven ID Preserving Image Generation Network (AAAI 2025 · similarity 0.85)
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.
