Summary of Meet Jeanie: a Similarity Measure For 3d Skeleton Sequences Via Temporal-viewpoint Alignment, by Lei Wang and Jun Liu and Liang Zheng and Tom Gedeon and Piotr Koniusz
Meet JEANIE: a Similarity Measure for 3D Skeleton Sequences via Temporal-Viewpoint Alignment
by Lei Wang, Jun Liu, Liang Zheng, Tom Gedeon, Piotr Koniusz
First submitted to arxiv on: 7 Feb 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a novel approach to address the issue of nuisance variations in video sequences, specifically speed of actions, temporal locations, and subjects’ poses. The authors introduce Joint tEmporal and cAmera viewpoiNt alIgnmEnt (JEANIE) for sequence pairs, which enables effective comparison and evaluation of two sets of frames or similarity of two sequences. The paper focuses on 3D skeleton sequences, where camera and subjects’ poses can be easily manipulated in 3D. The authors evaluate JEANIE on skeletal Few-shot Action Recognition (FSAR), a task that requires matching temporal blocks of support-query sequence pairs to factor out nuisance variations. JEANIE creates multiple views of a query sequence by simulating various camera locations, which are then matched with the support sequence using Dynamic Time Warping (DTW). The approach selects the smallest distance among matching paths with different temporal-viewpoint warping patterns, an advantage over DTW. The authors also propose an unsupervised FSAR method using JEANIE as a distance measure. The results demonstrate state-of-the-art performance on NTU-60, NTU-120, Kinetics-skeleton, and UWA3D Multiview Activity II for both supervised and unsupervised FSAR tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps solve a big problem in video analysis by making it easier to compare and understand different videos. The authors created a new way called JEANIE that can match up different parts of two videos, even if they were taken from different angles or with different actions. This is helpful for recognizing and understanding actions in videos. JEANIE works by creating many versions of a video from different camera angles, then matching those versions to another video using a special method called Dynamic Time Warping (DTW). JEANIE’s approach is better than DTW because it can find the best match even when there are small differences between the videos. The authors also came up with an unsupervised way of doing this that doesn’t need any labels or training. |
Keywords
* Artificial intelligence * Alignment * Few shot * Supervised * Unsupervised
