Research map / Continual learning / Online and replay-based
ForgetMeNot: Memory-Aware Forensic Facial Sketch Matching
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use the MTL-GPR 24-hour-to-1-hour memory model for matching | Forensic sketch matching · FSMD good-quality sketches, 10,030 mugshots | Rank 1 accuracy: 33 vs 17 | Helped |
| Add predicted facial attributes to MTL-GPR Late-Mem (ablation) | Forensic sketch matching · FSMD good-quality sketches, 10,030 mugshots | Rank 1 accuracy: 33 -> 38 | Helped |
| Add predicted facial attributes to MTL-GPR Late-Mem (ablation) | Forensic sketch matching · FSMD good-quality sketches, 10,030 mugshots | Rank 10 accuracy: 33 -> 42 | Helped |
| Add predicted facial attributes to MTL-GPR Late-Mem (ablation) | Forensic sketch matching · FSMD good-quality sketches, 10,030 mugshots | Rank 50 accuracy: 39 -> 45 | Helped |
| Use the MTL-GPR 24-hour-to-1-hour memory model | Forensic composite matching · Forensic composite database, 10,030 mugshots | Rank 1 accuracy: 14 vs 6 | Helped |
| Account for reconstruction variance when matching (ablation) | MTL-GPR memory-aware matching · MGDB, viewed sketch input | Rank 1 accuracy: 86 / 99 | Helped |
| Account for reconstruction variance when matching (ablation) | MTL-GPR memory-aware matching · MGDB, 24-hour sketch input | Rank 1 accuracy: 28 / 42 | Helped |
| Train related mapping tasks jointly with MTL-GPR rather than single-task GPR (ablation) | Memory-aware sketch reconstruction · MGDB, 24-hour-to-viewed task | Rank 1 accuracy: 57% vs 87% | Helped |
| Apply multiple single-task GPR corrections sequentially instead of one memory mapping (ablation) | Forensic sketch matching · FSMD forensic sketches | Rank 1 accuracy: 54; 28; 20; 13; 56; 39; 16; 16 | Hurt |
| Use a 24-hour-to-1-hour memory correction rather than other tested single memory tasks (ablation) | Forensic sketch matching · FSMD good-quality sketches, 10,030 mugshots | Rank 1 accuracy: 33 vs 23 | Helped |
| Use a learned reconstruction rather than direct HOG matching | Forensic composite matching · Forensic composite database, 10,030 mugshots | Rank 1 accuracy: 6 -> 14 | Helped |
| Learn a projection to reconstruct sketches | MGDB sketches · MGDB test split, external region | RMSE: 0.20 ± 0.013 -> 0.16±0.025 | Helped |
Closest papers by meaning
- Sketch Down the FLOPs: Towards Efficient Networks for Human Sketch (CVPR 2025 · similarity 0.85)
- Optimal Transport-based Labor-free Text Prompt Modeling for Sketch Re-identification (NeurIPS 2024 · similarity 0.84)
- How to Handle Sketch-Abstraction in Sketch-Based Image Retrieval? (CVPR 2024 · similarity 0.84)
- Forensics Adapter: Adapting CLIP for Generalizable Face Forgery Detection (CVPR 2025 · similarity 0.84)
- Stacking Brick by Brick: Aligned Feature Isolation for Incremental Face Forgery Detection (CVPR 2025 · similarity 0.83)
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · similarity 0.83)
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.