Research map / Video / Action recognition
Learning a Condensed Frame for Memory-Efficient Video Class-Incremental Learning
Read the paperCheck 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 Instance-Specific Prompt to a Frame Condensing frame (ablation) | FrameMaker · UCF101 10 steps | CNN accuracy: 72.29 -> 72.93 | Helped |
| add Instance-Specific Prompt to a Frame Condensing frame (ablation) | FrameMaker · HMDB51 5 steps | CNN accuracy: 42.18 -> 43.39 | Helped |
| add Instance-Specific Prompt to random frame (ablation) | single stored frame · UCF101 10 steps | CNN accuracy: 68.64 -> 70.71 | Helped |
| use Frame Condensing instead of averaging frames (ablation) | FrameMaker · UCF101 10 steps | CNN accuracy: 70.82 -> 72.29 | Helped |
| replace storing all frames with one condensed frame plus prompt (ablation) | FrameMaker · HMDB51 5 steps | CNN accuracy: 43.38 -> 43.39 | No clear effect |
| add condensing consistency loss Lc_f on top of Lc_ce (ablation) | Frame Condensing · UCF101 10 steps | CNN accuracy: 71.96 -> 72.29 | Helped |
| add cross entropy loss Lc_ce on top of Lc_f (ablation) | Frame Condensing · UCF101 10 steps | CNN accuracy: 71.35 -> 72.29 | Helped |
| add auxiliary condensing losses Lc to the prompt objective (ablation) | Instance-Specific Prompt · UCF101 10 steps | CNN accuracy: 71.83 -> 72.93 | Helped |
| add Lp_ce to prompt objective with Lp_f (ablation) | Instance-Specific Prompt · UCF101 10 steps | CNN accuracy: 72.35 -> 72.93 | Helped |
| increase condensed frames T from 8 to 16 (ablation) | Frame Condensing · UCF101 10 steps | CNN accuracy: 72.93 -> 72.42 | Hurt |
| instance-specific instead of task-specific prompt on frame (ablation) | prompt · HMDB51 5 steps | CNN accuracy: 41.16 -> 43.39 | Helped |
| frame-based prompt instead of feature-based prompt (ablation) | Instance-Specific Prompt · HMDB51 5 steps | CNN accuracy: 42.19 -> 43.39 | Helped |
Closest papers by meaning
- vCLIMB: A Novel Video Class Incremental Learning Benchmark (CVPR 2022 · similarity 0.87)
- StPR: Spatiotemporal Preservation and Routing for Exemplar-Free Video Class-Incremental Learning (ICLR 2026 · similarity 0.86)
- Learning Conditional Space-Time Prompt Distributions for Video Class-Incremental Learning (CVPR 2025 · similarity 0.85)
- Hypercorrelation Evolution for Video Class-Incremental Learning (AAAI 2024 · similarity 0.84)
- Coherent Temporal Synthesis for Incremental Action Segmentation (CVPR 2024 · similarity 0.84)
- Visual Representation Learning with Stochastic Frame Prediction (ICML 2024 · 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.