Research map / Detection and segmentation / CNN features
Real-Time Facial Segmentation and Performance Capture from RGB Input
Shunsuke Saito, Tianye Li, Hao Li
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 |
|---|---|---|---|
| use two-stream deconvolution network (FCN + DeconvNet) instead of FCN-8s alone | segmentation network · COFW segmentation | mean IOU: 0.739 -> 0.833 | Improved |
| use two-stream deconvolution network instead of DeconvNet alone | segmentation network · COFW segmentation | mean IOU: 0.821 -> 0.833 | Improved |
| use shared-convolution two-stream network instead of naive ensemble EDeconvNet | segmentation network · COFW segmentation | FPS: 20.45 -> 43.27 | Improved |
| use shared-convolution two-stream network instead of naive ensemble EDeconvNet | segmentation network · COFW segmentation | mean IOU: 0.835 -> 0.833 | No clear effect |
| add graph cut refinement to the segmentation output (ablation) | segmentation network · COFW segmentation | mean IOU: 0.833 -> 0.839 | Improved |
| add graph cut refinement to the segmentation output (ablation) | segmentation network · COFW segmentation | global accuracy: 0.882 -> 0.887 | Improved |
| add graph cut refinement to the segmentation output (ablation) | segmentation network · COFW segmentation | ave(face) recall: 0.929 -> 0.927 | Hurt |
| compare against RPP | RPP · COFW segmentation | mean IOU: 0.724 -> 0.833 | Improved |
| compare against SAPM | SAPM · COFW segmentation | mean IOU: 0.835 -> 0.833 | No clear effect |
| compare against Structured Forest | Structured Forest · COFW segmentation | global accuracy: 0.839 -> 0.882 | Improved |
| add hand compositing to segmentation training data (ablation) | segmentation network · Hand occlusion probability maps | probability map accuracy | Improved |
| train DDE regressor on segmented (masked) faces (ablation) | DDE regression · Synthetic box occlusion tracking | tracking error | Improved |
Closest papers by meaning
- Learning Dense Facial Correspondences in Unconstrained Images (ICCV 2017 · similarity 0.88)
- Towards Learning Structure via Consensus for Face Segmentation and Parsing (CVPR 2020 · similarity 0.88)
- DeepFaceFlow: In-the-Wild Dense 3D Facial Motion Estimation (CVPR 2020 · similarity 0.87)
- Generating 3D Faces using Convolutional Mesh Autoencoders (ECCV 2018 · similarity 0.87)
- Pre-training Strategies and Datasets for Facial Representation Learning (ECCV 2022 · similarity 0.87)
- 3D Face Tracking from 2D Video through Iterative Dense UV to Image Flow (CVPR 2024 · similarity 0.87)
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.
