Research map / Label-efficient and robust learning / Recognition and long-tail
MSML: Enhancing Occlusion-Robustness by Multi-Scale Segmentation-Based Mask Learning for Face Recognition
Ge Yuan, Huicheng Zheng, Jiayu Dong
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 multi-scale segmentation-based masking to ArcFace-18 | ArcFace-18 · MFV | verification accuracy (%): 95.40 -> 98.90 | Improved |
| Add multi-scale segmentation-based masking to ArcFace-18 | ArcFace-18 · MFV | TAR(@FAR=1e-3) (%): 81.10 -> 91.93 | Improved |
| Add multi-scale segmentation-based masking to ArcFace-18 | ArcFace-18 · MF1occ | identification accuracy (%): 64.50 -> 68.33 | Improved |
| Add multi-scale segmentation-based masking to LightCNN-29 | LightCNN-29 · AR Protocol 2, Scarf | identification accuracy (%): 96.76 -> 99.37 | Improved |
| Use a soft sigmoid mask instead of binarization (ablation) | MSML v1 configuration on L29 · LFW block occlusion 1%-40% | verification accuracy (%): 93.00 -> 94.43 | Improved |
| Use concatenated facial features and segmentation representations to generate masks (ablation) | MSML on L29 · LFW block occlusion 1%-40% | verification accuracy (%): 94.43 -> 95.27 | Improved |
| Add two ResBlocks to the mask operator (ablation) | MSML on L29 · LFW block occlusion 1%-40% | verification accuracy (%): 95.87 -> 96.30 | Improved |
| Remove the layer 1 masking operator (ablation) | MSML on L29 · LFW block occlusion 1%-40% | verification accuracy (%): 96.30 -> 91.73 | Hurt |
| Remove the layer 2 masking operator (ablation) | MSML on L29 · LFW block occlusion 1%-40% | verification accuracy (%): 96.30 -> 95.40 | Hurt |
| Remove the layer 3 masking operator (ablation) | MSML on L29 · LFW block occlusion 1%-40% | verification accuracy (%): 96.30 -> 95.35 | Hurt |
| Remove the layer 4 masking operator (ablation) | MSML on L29 · LFW block occlusion 1%-40% | verification accuracy (%): 96.30 -> 90.27 | Hurt |
| Replace the OSB ResNet-18 encoder with ResNet-34 (ablation) | MSML with ArcFace-18 · MFV | verification accuracy (%): 98.90 -> 98.43 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CosFace — CosFace: Large Margin Cosine Loss for Deep Face Recognition (CVPR 2018) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Occlusion Robust Face Recognition Based on Mask Learning With Pairwise Differential Siamese Network (ICCV 2019 · similarity 0.88)
- End2End Occluded Face Recognition by Masking Corrupted Features (TPAMI 2021 · similarity 0.87)
- Robust Facial Landmark Detection via Occlusion-Adaptive Deep Networks (CVPR 2019 · similarity 0.86)
- Towards Learning Structure via Consensus for Face Segmentation and Parsing (CVPR 2020 · similarity 0.84)
- Face Mask Extraction in Video Sequence (IJCV 2018 · similarity 0.84)
- Masked Face Recognition with Generative-to-Discriminative Representations (ICML 2024 · 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.
