Research map / Video / Video segmentation and tracking
SumGraph: Video Summarization via Recursive Graph Modeling
Jungin Park, Jiyoung Lee, Ig-Jae Kim, Kwanghoon Sohn
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 sparsity loss to classification loss (ablation) | SumGraph · TVSum (standard supervised) | F-score: 62.8 -> 63.6 | Improved |
| add diversity and reconstruction losses to classification loss (ablation) | SumGraph · TVSum (standard supervised) | F-score: 62.8 -> 63.5 | Improved |
| use all four losses instead of classification only (ablation) | SumGraph · TVSum (standard supervised) | F-score: 62.8 -> 63.9 | Improved |
| increase number of recursive graph refinement iterations (ablation) | SumGraph · TVSum (standard supervised) | F-score | Improved |
| augment training data with YouTube and OVP (and other dataset) | SumGraph · SumMe (supervised) | F-score: 51.4 -> 52.9 | Improved |
| augment training data with other datasets | SumGraph · TVSum (supervised) | F-score: 63.9 -> 65.8 | Improved |
| transfer setting instead of standard training (train on other datasets only) | SumGraph · SumMe (supervised) | F-score: 51.4 -> 48.7 | Hurt |
| train without ground-truth summaries (unsupervised losses) | SumGraph · SumMe (standard) | F-score: 51.4 -> 49.8 | Hurt |
| train without ground-truth summaries (unsupervised losses) | SumGraph · TVSum (standard) | F-score: 63.9 -> 59.3 | Hurt |
| recursively refined graph vs prior frame-wise baselines (SumGraph vs Fajtl et al.) | Fajtl et al. · SumMe (standard, supervised) | F-score: 49.7 -> 51.4 | Improved |
Later papers that compared with this method
4 later accepted papers report SumGraph as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Language-Guided Graph Representation Learning for Video Summarization (TPAMI 2025 · better on 3 of 4 shared results)
- Video Joint Modelling Based on Hierarchical Transformer for Co-summarization (TPAMI 2022 · better on 1 of 2 shared results)
- CLIP-It! Language-Guided Video Summarization (NeurIPS 2021 · better on 12 of 12 shared results)
- Multiple Pairwise Ranking Networks for Personalized Video Summarization (ICCV 2021 · better on 1 of 1 shared result)
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
- Language-guided Recursive Spatiotemporal Graph Modeling for Video Summarization (IJCV 2025 · similarity 0.96)
- Language-Guided Graph Representation Learning for Video Summarization (TPAMI 2025 · similarity 0.92)
- Reconstructive Sequence-Graph Network for Video Summarization (TPAMI 2021 · similarity 0.91)
- Video Summarization Using Fully Convolutional Sequence Networks (ECCV 2018 · similarity 0.90)
- Contextual Correspondence Matters: Bidirectional Graph Matching for Video Summarization (ECCV 2024 · similarity 0.90)
- Video Summarization with Large Language Models (CVPR 2025 · 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.
