Review Details (Adapted from CVPR and NIPS review instructions)
Your name, ID, Email
Last updated: Boqing Gong < http://crcv.ucf.edu/people/faculty/Gong/>
Paper / Gong, Boqing, Wei-Lun Chao, Kristen Grauman, and Fei Sha. "Diverse Sequential Subset Selection for Supervised Video Summarization." In Advances in Neural Information Processing Systems, pp. 2069-2077. 2014.Q1. Paper summary
This paper presents a framework for supervised video summarization. The proposed method includes a new Sequential Determinantal Point Process (seqDPP) model, which learns to select the best subsets of the raw input data as components of the output summarization. Unlike the most common approaches, that focus on unsupervised summarization by selecting most important/representative frames and features, this paper attempts to directly learn from human-created summaries.Q2. Paper Strengths. Please discuss the positive aspects of the paper. Be sure to comment on the paper's novelty, technical correctness, clarity and experimental evaluation. Notice that different papers may need different levels of evaluation: a theoretical paper may need no experiments, while a paper presenting a new approach to a known problem may require thorough comparisons to existing methods. Also, please make sure to justify your comments in great detail. For example, if you think the paper is novel, not only say so, but also explain in detail why you think this is the case.
The technical formulation of the method seems sound and correct. The new model seqDPP extends DPP by incorporating sequential dependencies. The model, learning and inference formulations are updated to account for the extension.The performance evaluation is adequate, including quantitative results (Precision, Recall and F-Score) from 3 datasets (youtube, OVP, kodak consumer). The results are good and mostly outperform the related work.
The readability of the document is good.
In terms of originality, the seqDPP model is an extension of DPP which is used to summarize text data. The extension is sound to tackle the problem of video summarization. Also, the authors use a novel way to derive ground truth, by combining multiple human-created summaries into a unique ‘oracle’ summary, which is used for learning and evaluation.
Q3. Paper Weaknesses. Please discuss the negative aspects of the paper: lack of novelty or clarity, technical errors, insufficient experimental evaluation, etc. Justify your comments in great detail. If you think the paper is not novel, explain why and give a reference to prior work. Keep in mind that novelty can take a number of forms; a paper may be novel in terms of the method, the problem, the theory, analysis for an existing problem, or the empirical evaluation. If you think there is an error in the paper, explain in detail why it is an error. If you think the experimental evaluation is insufficient, remember that theoretical results/ideas are essential to CVPR and that a theoretical paper need not have experiments. It is *not* okay to reject a paper because it did not outperform other existing algorithms, especially if the theory is novel and interesting.
Q4. Overall rating
0: Top 10% of the papers I have read, an excellent paper, a strong accept.
- I will fight for acceptance. I will consider not reviewing papers for XX if this is rejected.
1: Top 50% of accepted NIPS papers, a very good paper, a clear accept.
- I vote and argue for acceptance.
2: Good paper, accept.
- I vote for acceptance, although would not be upset if it were rejected.
3: Marginally above the acceptance threshold.
- I tend to vote for accepting it, but leaving it out of the program would be no great loss.
4: Marginally below the acceptance threshold.
- I tend to vote for rejecting it, but having it in the program would not be that bad.
5: An OK paper, but not good enough. A rejection.
- I vote for rejecting it, although would not be upset if it were accepted.
Q5. Explain your rating: how you weigh the strengths and weaknesses
The authors present a novel model, with an appropriate evaluation getting good results which mostly outperforms related work.Q6. Future directions (New solutions, new solutions extending or inspired by the paper’s solution, open problems, other problems that could benefit from the paper).