Research map / Video / Video segmentation and tracking
Face Mask Extraction in Video Sequence
Yujiang Wang, Bingnan Luo, Jie Shen, Maja Pantić
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 5-frame Gaussian temporal smoothing to the predictions (ablation) | FCN-ResNet50 · 300VW-Mask test | mean IoU (%): 54.50 -> 54.21 | No clear effect |
| Add a ConvLSTM layer (other layers frozen, Adam) (ablation) | FCN-ResNet50 · 300VW-Mask test | mean IoU (%): 54.50 -> 55.53 | Improved |
| Use Adam instead of RMSprop for the ConvLSTM layer (ablation) | ConvLSTM-FCN · 300VW-Mask test | mean IoU (%): 54.93 -> 55.53 | Improved |
| Replace cross-entropy with IoU loss (ablation) | ConvLSTM-FCN (frozen) · 300VW-Mask test | mean IoU (%): 55.53 -> 57.22 | Improved |
| Replace cross-entropy with Segmentation Loss (g=0) (ablation) | ConvLSTM-FCN (frozen) · 300VW-Mask test | mean IoU (%): 55.53 -> 59.04 | Improved |
| Use Segmentation Loss (g=0) instead of IoU loss (ablation) | ConvLSTM-FCN (frozen) · 300VW-Mask test | mean IoU (%): 57.22 -> 59.04 | Improved |
| Unfreeze other layers with reduced learning rate (gamma=0.05) (ablation) | ConvLSTM-FCN with SegLoss · 300VW-Mask test | mean IoU (%): 59.04 -> 60.04 | Improved |
| Vary gamma from 0.01 to 0.1 for the primary model (ablation) | Primary ConvLSTM-FCN · 300VW-Mask test | mean IoU (%): 60.35, 60.96, 60.04, 60.07 | No clear effect |
| Add ConvLSTM with gamma=0.1 to the eye model (ablation) | Eye baseline-FCN · Eye sub-dataset | Eyes IoU (%): 54.29 -> 51.82 | Hurt |
| Add ConvLSTM with gamma=0.02 to the eye model (ablation) | Eye baseline-FCN · Eye sub-dataset | Eyes IoU (%): 54.29 -> 59.01 | Improved |
| Add ConvLSTM to the mouth model (gamma=0.02) (ablation) | Mouth baseline-FCN · Mouth sub-dataset | mean IoU (%): 49.77 -> 52.17 | Improved |
| Add the two additional eye and mouth models to the primary model (ablation) | Primary model · 300VW-Mask test | mean IoU (%): 60.04 -> 63.76 | Improved |
Closest papers by meaning
- Parameter Efficient Local Implicit Image Function Network for Face Segmentation (CVPR 2023 · similarity 0.86)
- Mask Propagation for Efficient Video Semantic Segmentation (NeurIPS 2023 · similarity 0.86)
- FML: Face Model Learning From Videos (CVPR 2019 · similarity 0.86)
- Real-Time Facial Segmentation and Performance Capture from RGB Input (ECCV 2016 · similarity 0.86)
- Unconstrained Still/Video-Based Face Verification with Deep Convolutional Neural Networks (IJCV 2017 · similarity 0.86)
- Learning Detailed Face Reconstruction From a Single Image (CVPR 2017 · similarity 0.85)
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.
