Research map / 3D vision / Depth and stereo
CriSp: Leveraging Tread Depth Maps for Enhanced Crime-Scene Shoeprint Matching
Samia Shafique, Shu Kong, Charless Fowlkes
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 crime-scene data augmentation (occlusion, erasure, noise) to training (ablation) | CriSp · val-FID | mAP@100: 0.0174 -> 0.2071 | Improved |
| add data augmentation to ZSE-SBIR (ablation) | ZSE-SBIR · val-FID | hit@100: 0.0943 -> 0.4528 | Improved |
| match against tread depth maps only instead of RGB, print or all three (ablation) | database image configuration · FID-clean | mAP@100: 0.066 (RGB) -> 0.203 (depth) | Improved |
| combine RGB, depth and print in database images (ablation) | database image configuration · FID-clean | mAP@100: 0.203 -> 0.057 | Hurt |
| add occlusion, erasure and noise augmentations together (ablation) | ResNet50 with SupCon · FID-crime | hit@100: 0.009 -> 0.226 | Improved |
| add occlusion augmentation alone (ablation) | ResNet50 with SupCon · FID-crime | hit@100: 0.009 -> 0.019 | No clear effect |
| use spatial features by removing last pooling (ablation) | ResNet50 · val-FID FID-crime | hit@100: 0.2264 -> 0.3585 | Improved |
| add spatial feature masking (ablation) | ResNet50 · val-FID FID-crime | hit@100: 0.3585 -> 0.4245 | Improved |
| add query print masking during training (ablation) | CriSp Enc · val-FID FID-crime | hit@100: 0.4528 -> 0.5472 | Improved |
| train prior methods with simulated partial-print masks | FIRe, SupCon, ZSE-SBIR · val-FID | mAP@100: FIRe 0.0398 -> 0.0030; SupCon 0.0096 -> 0.0001 | Hurt |
| match MCNCC against clean shoeprints instead of tread depth maps | MCNCC · val-FID | hit@100: 0.0849 -> 0.0283 | Hurt |
Closest papers by meaning
- Cross-Domain Image Matching with Deep Feature Maps (IJCV 2019 · similarity 0.88)
- ForensicHub: A Unified Benchmark & Codebase for All-Domain Fake Image Detection and Localization (NeurIPS 2025 · similarity 0.82)
- ShoeFit: A New Dataset and Dual-image-stream DiT Framework for Virtual Footwear Try-On (NeurIPS 2025 · similarity 0.80)
- Towards Explainable and Unprecedented Accuracy in Matching Challenging Finger Crease Patterns (CVPR 2025 · similarity 0.79)
- Forensic Self-Descriptions Are All You Need for Zero-Shot Detection, Open-Set Source Attribution, and Clustering of AI-generated Images (CVPR 2025 · similarity 0.79)
- Locate-Then-Examine: Grounded Region Reasoning Improves Detection of AI-Generated Images (CVPR 2026 · similarity 0.79)
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.
