Research map / Detection and segmentation / Open-set and few-shot detection
High-Quality Correspondence and Segmentation Estimation for Dual-Lens Smart-Phone Portraits
Xiaoyong Shen, Hongyun Gao, Xin Tao, Chao Zhou, Jiaya Jia
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 joint correspondence refinement to the segmentation CRF (vs updating segmentation only) (ablation) | Ours · Dual-lens portrait segmentation (200 test pairs) | mean IoU (%): 84.32 -> 88.33 | Improved |
| add joint segmentation refinement to the correspondence CRF (vs updating correspondence only) (ablation) | Ours · Dual-lens portrait correspondence (200 test pairs) | AEPE: 8.03 -> 5.29 | Improved |
| add joint segmentation refinement to the correspondence CRF (vs updating correspondence only) (ablation) | Ours · Dual-lens portrait correspondence (200 test pairs) | AAE: 7.45 -> 5.91 | Improved |
| add regional correspondence (vs 500 uniform constant displacement labels) (ablation) | Ours · Regional correspondence study | AEPE: 6.45 -> 5.29 | Improved |
| add regional correspondence (vs 500 uniform constant displacement labels) (ablation) | Ours · Regional correspondence study | running time (s): 186.3 -> 16.63 | Improved |
| fine-tune 2-class FCN on portrait data | FCN · Dual-lens portrait segmentation (200 test pairs) | mean IoU (%): 79.51 -> 83.90 | Improved |
| fine-tune 2-class DeepLab on portrait data | DeepLab · Dual-lens portrait segmentation (200 test pairs) | mean IoU (%): 80.09 -> 84.01 | Improved |
| fine-tune 2-class CRFasRNN on portrait data | CRFasRNN · Dual-lens portrait segmentation (200 test pairs) | mean IoU (%): 80.23 -> 84.19 | Improved |
| update segmentation only with the CRF on top of fine-tuned FCN | FCN-portrait · Dual-lens portrait segmentation (200 test pairs) | mean IoU (%): 83.90 -> 84.32 | No clear effect |
| replace fully connected CRF pairwise term with MRF neighbourhood term (ablation) | Ours · Qualitative arm region example | visual quality | Hurt |
Closest papers by meaning
- Joint Recovery of Dense Correspondence and Cosegmentation in Two Images (CVPR 2016 · similarity 0.82)
- DualRC: A Dual-Resolution Learning Framework With Neighbourhood Consensus for Visual Correspondences (TPAMI 2023 · similarity 0.82)
- Learning Probabilistic Coordinate Fields for Robust Correspondences (TPAMI 2023 · similarity 0.81)
- RUBIK: A Structured Benchmark for Image Matching across Geometric Challenges (CVPR 2025 · similarity 0.81)
- Dual-Resolution Correspondence Networks (NeurIPS 2020 · similarity 0.81)
- Unconstrained Motion Deblurring for Dual-Lens Cameras (ICCV 2019 · similarity 0.81)
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.
