Research map / Label-efficient and robust learning / Recognition and long-tail
Deep Perceptual Mapping for Cross-Modal Face Recognition
M. Saquib Sarfraz, Rainer Stiefelhagen
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 DPM mapping to baseline features (ablation) | SIFT block features · UND-X1 thermal-visible (1 visible/subject gallery) | Rank-1 (%): 30.36 -> 55.36 | Improved |
| add DPM mapping to baseline features (ablation) | SIFT block features · Carl LWIR-visible (1 visible/subject gallery) | Rank-1 (%): 26.50 -> 56.33 | Improved |
| add DPM mapping to baseline features (ablation) | SIFT block features · NVESD MWIR-visible (1 visible/subject gallery) | Rank-1 (%): 72.00 -> 86.00 | Improved |
| add DPM mapping to baseline features (ablation) | SIFT block features · NVESD LWIR-visible (1 visible/subject gallery) | Rank-1 (%): 61.00 -> 82.67 | Improved |
| use DPM deep mapping instead of linear PLS mapping (ablation) | SIFT block features · UND-X1 thermal-visible (1 visible/subject gallery) | Rank-1 (%): 44.75 -> 55.36 | Improved |
| use DPM deep mapping instead of linear PLS mapping (ablation) | SIFT block features · Carl LWIR-visible (1 visible/subject gallery) | Rank-1 (%): 31.75 -> 56.33 | Improved |
| use DPM deep mapping instead of linear PLS mapping (ablation) | SIFT block features · NVESD LWIR-visible (all visible/subject gallery) | Rank-1 (%): 93.66 -> 97.33 | Improved |
| add second hidden layer to DPM (two vs one hidden layer) (ablation) | DPM · UND-X1 thermal-visible (1 visible/subject gallery) | Rank-1 (%): 50.67 -> 55.36 | Improved |
| add second hidden layer to DPM (two vs one hidden layer) (ablation) | DPM · Carl LWIR-visible (1 visible/subject gallery) | Rank-1 (%): 50.42 -> 56.33 | Improved |
| add second hidden layer to DPM (two vs one hidden layer) (ablation) | DPM · NVESD MWIR-visible (1 visible/subject gallery) | Rank-1 (%): 86.66 -> 86.00 | No clear effect |
| add second hidden layer to DPM (two vs one hidden layer) (ablation) | DPM · NVESD LWIR-visible (1 visible/subject gallery) | Rank-1 (%): 83.67 -> 82.67 | No clear effect |
| use SIFT instead of HOG descriptors (ablation) | DPM · thermal-visible datasets | Rank-1 (%): HOG produces 2-3% inferior results than SIFT | Improved |
Closest papers by meaning
- Hierarchical Discriminative Learning for Visible Thermal Person Re-Identification (AAAI 2018 · similarity 0.87)
- CM-NAS: Cross-Modality Neural Architecture Search for Visible-Infrared Person Re-Identification (ICCV 2021 · similarity 0.86)
- A Comprehensive Database for Benchmarking Imaging Systems (TPAMI 2018 · similarity 0.85)
- Disentangled Variational Representation for Heterogeneous Face Recognition (AAAI 2019 · similarity 0.85)
- Cross-Modality Transformer for Visible-Infrared Person Re-identification (ECCV 2022 · similarity 0.85)
- Pseudo Visible Feature Fine-Grained Fusion for Thermal Object Detection (CVPR 2025 · similarity 0.84)
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.
