Research map / Video / Video segmentation and tracking
Exploring Decoupled Spatio-Temporal Consistency Learning and Self-Prompting Evolution for Self-Supervised Tracking
Yaozong Zheng, Bineng Zhong, Qihua Liang, Ning Li, Haiying Xia, Shuxiang Song, Rongrong Ji
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 decoupled training framework to contrastive baseline (ablation) | SSTrack · GOT-10k | AO: 41.8 -> 68.6 | Improved |
| add spatio-temporal context (ablation) | SSTrack · GOT-10k | AO: 68.6 -> 70.8 | Improved |
| add instance contrastive loss (ablation) | SSTrack · GOT-10k | AO: 70.8 -> 72.4 | Improved |
| replace MAE pre-training with DropMAE (ablation) | SSTrack · GOT-10k | AO: 62.7 -> 68.6 | Improved |
| add shear, blur and LSJ augmentations (ablation) | SSTrack · GOT-10k | AO: 68.6 -> 70.3 | Improved |
| add weak-to-strong training framework (ablation) | SSTrack++ · LaSOT | AUC: 65.9 -> 66.3 | Improved |
| add self-prompting evolution (SPE) (ablation) | SSTrack++ · LaSOT | AUC: 65.3 -> 66.3 | Improved |
| use query association instead of SPE (ablation) | SSTrack++ · LaSOT | AUC: 66.3 -> 65.0 | Hurt |
| increase self-prompts from 8 to 32 (ablation) | SSTrack++ · LaSOT | AUC: 66.3 -> 61.5 | Hurt |
| increase self-prompts from 4 to 8 (ablation) | SSTrack++ · LaSOT | AUC: 65.4 -> 66.3 | Improved |
| use token filtering ratio of 70% (ablation) | SSTrack++ · LaSOT | AUC | Mixed |
| use SSTrack++-384 instead of SSTrack-384 (add SPE and weak-to-strong) | SSTrack++-384 · LaSOT | AUC: 65.9 -> 66.9 | Improved |
Closest papers by meaning
- Decoupled Spatio-Temporal Consistency Learning for Self-Supervised Tracking (AAAI 2025 · similarity 0.96)
- SOVTrack: Open-Vocabulary Multi-Object Tracking with Self-Supervised Pseudo Labeling and Feature Distillation (ECCV 2026 · similarity 0.88)
- Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision (NeurIPS 2025 · similarity 0.87)
- Autoregressive Queries for Adaptive Tracking with Spatio-Temporal Transformers (CVPR 2024 · similarity 0.86)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · similarity 0.86)
- Sequential Fusion Based Multi-Granularity Consistency for Space-Time Transformer Tracking (AAAI 2024 · 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.
