Research map / Video / Video segmentation and tracking
Video Region Annotation with Sparse Bounding Boxes
Yuzheng Xu, Yang Wu, Nur Sabrina binti Zuraimi, Shohei Nobuhara, Ko Nishino
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 motion features (ablation) | VGCN-basic · Synthia | mIoU (%): 66.41 -> 66.92 | Improved |
| add full local connection (ablation) | VGCN-basic · Synthia | mIoU (%): 66.41 -> 67.12 | Improved |
| add both motion features and full-connection (ablation) | VGCN-basic · Synthia | mIoU (%): 66.41 -> 67.80 | Improved |
| use volumetric graph with temporal connections instead of frame-wise SGCN | SGCN · Synthia | mIoU (%): 65.19 -> 67.80 | Improved |
| add temporal smoothing with B-spline to SGCN (ablation) | SGCN · Synthia | mIoU (%): 65.19 -> 65.37 | No clear effect |
| add motion features to VGCN-basic (ablation) | VGCN-basic · KITTI-MOTS (direct transfer, Ped) | mIoU (%): 63.3 -> 61.7 | Hurt |
| add full-connection to VGCN-basic (ablation) | VGCN-basic · KITTI-MOTS (fine-tuned, Ped) | mIoU (%): 63.1 -> 65.2 | Improved |
| fine-tune on KITTI-MOTS | VGCN-basic · KITTI-MOTS Car | mIoU (%): 70.3 -> 70.9 | No clear effect |
| fine-tune on KITTI-MOTS | VGCN · KITTI-MOTS Pedestrian | mIoU (%): 61.7 -> 65.1 | Improved |
| fine-tune on KITTI-MOTS | SGCN · KITTI-MOTS Car | mIoU (%): 58.2 -> 66.1 | Improved |
| use ground-truth boxes instead of interpolated boxes | VGCN · Synthia | mIoU (%): 67.80 -> 68.67 | Improved |
| use ground-truth boxes instead of interpolated boxes | VGCN · KITTI MOTS Ped | mIoU (%): 71.1% (6.0% up) | Improved |
Closest papers by meaning
- Generating Masks From Boxes by Mining Spatio-Temporal Consistencies in Videos (ICCV 2021 · similarity 0.87)
- G-TAD: Sub-Graph Localization for Temporal Action Detection (CVPR 2020 · similarity 0.86)
- Video Annotation for Visual Tracking via Selection and Refinement (ICCV 2021 · similarity 0.86)
- Track and Segment: An Iterative Unsupervised Approach for Video Object Proposals (CVPR 2016 · similarity 0.86)
- Bringing Image Scene Structure to Video via Frame-Clip Consistency of Object Tokens (NeurIPS 2022 · similarity 0.86)
- Local Compressed Video Stream Learning for Generic Event Boundary Detection (IJCV 2023 · similarity 0.86)
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.
