Research map / Video / Video-language understanding
Weakly Supervised Summarization of Web Videos
Rameswar Panda, Abir Das, Ziyan Wu, Jan Ernst, Amit K. Roy-Chowdhury
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 cross-dataset pre-training on UCF101 instead of training from scratch (ablation) | DeSumNet · CoSum | top-5 mAP relative to human (%): 71.4 -> 83.5 | Improved |
| add cross-dataset pre-training on UCF101 instead of training from scratch (ablation) | DeSumNet · TVSum | top-5 mAP relative to human (%): 66.7 -> 75.2 | Improved |
| add refined web-video model adaptation to pre-training (ablation) | DeSumNet · CoSum | top-5 mAP relative to human (%): 83.5 -> 87.7 | Improved |
| add refined web-video model adaptation to pre-training (ablation) | DeSumNet · TVSum | top-5 mAP relative to human (%): 75.2 -> 80.8 | Improved |
| add enhanced data augmentation on top of pre-training and adaptation (ablation) | DeSumNet · CoSum | top-5 mAP relative to human (%): 87.7 -> 88.5 | Improved |
| add enhanced data augmentation on top of pre-training and adaptation (ablation) | DeSumNet · TVSum | top-5 mAP relative to human (%): 80.8 -> 82.2 | Improved |
| add unrefined noisy web videos directly to scratch training (ablation) | DeSumNet · CoSum | top-5 mAP relative to human (%): 71.4 -> 76.3 | Improved |
| add unrefined noisy web videos directly to pre-trained training (ablation) | DeSumNet · CoSum | top-5 mAP relative to human (%): 83.5 -> 84.4 | Improved |
| add unrefined noisy web videos directly to pre-trained training (ablation) | DeSumNet · TVSum | top-5 mAP relative to human (%): 75.2 -> 77.3 | Improved |
| use refined web videos instead of unrefined noisy web videos after pre-training (ablation) | DeSumNet · CoSum | top-5 mAP relative to human (%): 84.4 -> 87.7 | Improved |
| enforce diversity by clustering segments before summary selection (ablation) | DeSumNet · CoSum | top-5 mAP: ~0.8% improvement | No clear effect |
| replace CVS with DeSumNet for time-lapse | time-lapse generation · CoSum | average human rating: 3.23 -> 4.03 | Improved |
Closest papers by meaning
- Weakly-supervised Video Summarization using Variational Encoder-Decoder and Web Prior (ECCV 2018 · similarity 0.91)
- Video Summarization by Learning From Unpaired Data (CVPR 2019 · similarity 0.90)
- Video Summarization with Long Short-Term Memory (ECCV 2016 · similarity 0.89)
- Retrospective Encoders for Video Summarization (ECCV 2018 · similarity 0.88)
- Video Summarization Using Fully Convolutional Sequence Networks (ECCV 2018 · similarity 0.88)
- Query-Focused Video Summarization: Dataset, Evaluation, and a Memory Network Based Approach (CVPR 2017 · similarity 0.88)
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.
