Research map / Continual learning / Class-incremental
Teacher Agent: A Knowledge Distillation-Free Framework for Rehearsal-Based Video Incremental Learning
Shengqin Jiang, Yaoyu Fang, Haokui Zhang, Qingshan Liu, Yuankai Qi, Yang Yang, Peng Wang
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 teacher agent (distillation-free) to baseline with sampler (ablation) | video incremental baseline · UCF101 | accuracy: 69.88 -> 85.53 | Improved |
| add unified sampler to low-resolution baseline (ablation) | low-res vCLIMB_TC · UCF101 | accuracy: 67.32 -> 69.88 | Improved |
| add unified sampler to low-resolution baseline (ablation) | low-res vCLIMB_TC · Something-Something V1 | accuracy: 8.23 -> 14.48 | Improved |
| reduce input resolution to 112x112 without other changes (ablation) | vCLIMB_TC · UCF101 | accuracy: 75.82 -> 67.32 | Hurt |
| use teacher-agent soft labels instead of one-hot ground truth (ablation) | Ours · UCF101 | accuracy: 82.74 -> 85.53 | Improved |
| freeze calibration parameters (Ours vs Ours(para)) (ablation) | Ours · UCF101 | accuracy: 85.40 -> 85.53 | Improved |
| use frozen parameters instead of random vectors for calibration input (ablation) | Ours · UCF101 | accuracy: 84.87 -> 85.53 | Improved |
| eliminate gradient of misclassified exemplars in baseline (ablation) | Baseline · UCF101 | accuracy: 67.32 -> 68.67 | Improved |
| add T-agent to image CIL model MEMO (ablation) | MEMO · CIFAR-100 10 tasks | accuracy: 57.52 -> 58.18 | Improved |
| add T-agent to WA on Tiny ImageNet 10 tasks (ablation) | WA · Tiny ImageNet 10 tasks | accuracy: 39.61 -> 40.62 | Improved |
| increase stored exemplars to four times | Ours · UCF101 | accuracy: 85.53 -> 95.56 | Improved |
| halve input resolution (reduction effect) for Ours vs benchmark (ablation) | Ours · UCF101 | accuracy: 88.55 -> 85.53 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- iCaRL — iCaRL: Incremental Classifier and Representation Learning (CVPR 2017) · better on 4 of 4 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
- Procedure Knowledge Decoupled Distillation Strategy for Procedure Planning in Instructional Videos (AAAI 2025 · similarity 0.86)
- MIND: Multi-Task Incremental Network Distillation (AAAI 2024 · similarity 0.86)
- Multi-Teacher Knowledge Distillation with Reinforcement Learning for Visual Recognition (AAAI 2025 · similarity 0.86)
- SLVMBench: Skill Learning from Video Memory (NeurIPS 2026 · similarity 0.85)
- Class Incremental Learning with Multi-Teacher Distillation (CVPR 2024 · similarity 0.85)
- Orchestrate Latent Expertise: Advancing Online Continual Learning with Multi-Level Supervision and Reverse Self-Distillation (CVPR 2024 · similarity 0.84)
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.
