Research map / Continual learning / Class-incremental
STSP: Spatial-Temporal Subspace Projection for Video Class-incremental Learning
Hao Cheng, SIYUAN YANG, Chong Wang, Joey Tianyi Zhou, Alex Kot, Bihan Wen
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 spatial-based gradient projection (SGP) in place of exemplar-based distillation (ablation) | STSP with TSC · UCF101 (5x10 stages) | accuracy: 71.50 -> 82.84 | Improved |
| add SGP to a model with no exemplars (ablation) | linear classifier · UCF101 (5x10 stages) | accuracy: 17.63 -> 77.66 | Improved |
| replace linear classifier with TSC (with SGP) (ablation) | STSP · UCF101 (5x10 stages) | accuracy: 77.66 -> 82.84 | Improved |
| replace linear classifier with TSC (with SGP) (ablation) | STSP · UCF101 (2x25 stages) | accuracy: 74.02 -> 79.25 | Improved |
| replace linear classifier with TSC (with exemplars) (ablation) | exemplar-based distillation · UCF101 (5x10 stages) | accuracy: 68.28 -> 71.50 | Improved |
| use TSC without SGP or exemplars (ablation) | linear classifier baseline · UCF101 (5x10 stages) | accuracy: 17.63 -> 8.78 | Hurt |
| use subspace basis size tau=8 instead of 16 (ablation) | STSP TSC · UCF101 (5x10 stages) | accuracy: 63.02 -> 82.84 | Improved |
| increase subspace basis size from 1 to 8 (ablation) | STSP TSC · UCF101 (5x10 stages) | accuracy: 66.75 -> 82.84 | Improved |
| STSP vs FrameMaker (exemplar-free vs exemplar-based) | STSP · Something-Something V2 (10x9 stages) | accuracy: 37.25 -> 69.68 | Improved |
| more incremental stages (2x25 instead of 5x10) | STSP · UCF101 | accuracy: 82.84 -> 79.25 | Hurt |
Later papers that compared with this method
1 later accepted paper reports STSP as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- StPR: Spatiotemporal Preservation and Routing for Exemplar-Free Video Class-Incremental Learning (ICLR 2026 · better on 5 of 5 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- StPR: Spatiotemporal Preservation and Routing for Exemplar-Free Video Class-Incremental Learning (ICLR 2026 · similarity 0.87)
- Hypercorrelation Evolution for Video Class-Incremental Learning (AAAI 2024 · similarity 0.86)
- Learning Conditional Space-Time Prompt Distributions for Video Class-Incremental Learning (CVPR 2025 · similarity 0.84)
- Expandable Subspace Ensemble for Pre-Trained Model-Based Class-Incremental Learning (CVPR 2024 · similarity 0.84)
- Tripartite Weight-Space Ensemble for Few-Shot Class-Incremental Learning (CVPR 2025 · similarity 0.83)
- State Space Prompting via Gathering and Spreading Spatio-Temporal Information for Video Understanding (NeurIPS 2025 · similarity 0.83)
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.
