Research map / Continual learning / Domain and prompt-based
Lifelong Unsupervised Domain Adaptive Person Re-Identification With Coordinated Anti-Forgetting and Adaptation
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 |
|---|---|---|---|
| Add coordinated data replay (CDR) to stage-wise UDA (ablation) | Stage-wise UDA · Dynamic target streams: Market1501, CUHK-SYSU, MSMT17 | R-1: 72.23 -> 73.12; 64.59 -> 69.23; 19.80 -> 23.56 | Helped |
| Add relational consistency learning (RCL) to CDR (ablation) | Base. + CDR · Dynamic target streams: Market1501, CUHK-SYSU, MSMT17 | R-1: 73.12 -> 75.21; 69.23 -> 75.38; 23.56 -> 28.67 | Helped |
| Add data replay (ablation) | Stage-wise UDA · Dynamic target streams | R-1: 72.23 -> 72.51; 64.59 -> 66.22; 19.80 -> 20.98 | Helped |
| Use task-oriented meta-train/meta-test splitting instead of data-based splitting (ablation) | CLUDA-ReID CDR · Market1501 stationary target scenario | R-1 | Helped |
| Use both old and new batches for KD in meta-test rather than only the old batch (ablation) | CLUDA-ReID CDR · Market1501 stationary target scenario | R-1 | Helped |
| Use both old and new batches for KD in meta-test rather than only the new batch (ablation) | CLUDA-ReID CDR · Market1501 stationary target scenario | R-1 | Helped |
| Replace ID-wise reservoir sampling with FIFO memory (ablation) | CLUDA-ReID · Stationary target scenario | R-1 | Hurt |
| Replace ID-wise reservoir sampling with instance-wise reservoir sampling (ablation) | CLUDA-ReID · Stationary target scenario | R-1 | Hurt |
| Replace the momentum-updated historical model with a snapshot from one past moment (ablation) | CLUDA-ReID · Stationary target scenario | R-1 | Hurt |
| Use only KL consistency instead of combined relational and KL consistency (ablation) | CLUDA-ReID RCL · Stationary target scenario | R-1 | Hurt |
| Use only relational consistency instead of combined relational and KL consistency (ablation) | CLUDA-ReID RCL · Stationary target scenario | R-1 | Hurt |
| Add KL consistency regularization to CDR (ablation) | Base. + CDR · Dynamic target streams | R-1: 73.12 -> 74.24; 69.23 -> 71.55; 23.56 -> 26.16 | Helped |
Closest papers by meaning
- Lifelong Person Re-Identification via Adaptive Knowledge Accumulation (CVPR 2021 · similarity 0.91)
- Lifelong Person Re-identification by Pseudo Task Knowledge Preservation (AAAI 2022 · similarity 0.89)
- DASK: Distribution Rehearsing via Adaptive Style Kernel Learning for Exemplar-Free Lifelong Person Re-Identification (AAAI 2025 · similarity 0.89)
- Learning Continual Compatible Representation for Re-indexing Free Lifelong Person Re-identification (CVPR 2024 · similarity 0.88)
- Distribution-aware Knowledge Prototyping for Non-exemplar Lifelong Person Re-identification (CVPR 2024 · similarity 0.86)
- LSTKC: Long Short-Term Knowledge Consolidation for Lifelong Person Re-identification (AAAI 2024 · 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.