Research map / Detection and segmentation / Open-set and few-shot detection
Learning Fast and Robust Target Models for Video Object Segmentation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace the offline-trained target encoder with an inference-learned discriminative target model (ablation) | VOS segmentation architecture · YouTube-VOS train validation split | J: 49.8 -> 58.3 | Helped |
| Use a shallow target model instead of first-frame fine-tuning (ablation) | VOS segmentation architecture · YouTube-VOS train validation split | J: 58.9 vs 58.3 | No clear effect |
| Add online target-model updates (ablation) | D-only target model · YouTube-VOS train validation split | J: 58.3 -> 59.6; +1.6% | Helped |
| Add the segmentation network to refine coarse target scores (ablation) | D-only target model · YouTube-VOS train validation split | J: 58.3 -> 67.9; +8.3% | Helped |
| Update the target model using refined segmentation predictions (ablation) | Ours - no update · YouTube-VOS train validation split | J: 67.9 -> 71.4 | Helped |
| Use ResNet-18 and fewer optimization steps instead of the full configuration (ablation) | Ours · YouTube-VOS 2018 validation | G: 72.1 -> 65.7 | Hurt |
| Use ResNet-18 and fewer optimization steps instead of the full configuration (ablation) | Ours · YouTube-VOS 2018 validation | FPS: 21.9 -> 41.3 | Helped |
| Train on all YouTube-VOS data rather than half the training split (ablation) | Ours · YouTube-VOS 2018 test-dev | G: 70.6 -> 72.1 | Helped |
| Train on half rather than 25% of YouTube-VOS (ablation) | Ours · YouTube-VOS 2018 test-dev | G: 66.7 -> 70.6 | Helped |
| Increase the initial sample-memory size from 1 to 5 (ablation) | Ours (yt) · YouTube-VOS validation split | J: 71.5 -> 71.4 | No clear effect |
| Increase the initial sample-memory size from 1 to 5 (ablation) | Ours (yt+dv) · DAVIS 2017 validation | J: 69.4 -> 73.8 | Helped |
Closest papers by meaning
- A Generative Appearance Model for End-To-End Video Object Segmentation (CVPR 2019 · similarity 0.92)
- Point-VOS: Pointing Up Video Object Segmentation (CVPR 2024 · similarity 0.91)
- Make One-Shot Video Object Segmentation Efficient Again (NeurIPS 2020 · similarity 0.91)
- Two-Shot Video Object Segmentation (CVPR 2023 · similarity 0.90)
- Breaking the "Object" in Video Object Segmentation (CVPR 2023 · similarity 0.90)
- Test-time Training for Matching-based Video Object Segmentation (NeurIPS 2023 · similarity 0.90)
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.