This code is provided by Jun Wan. It is used in the Chalearn one-shot learning gesture challenge (round 2). This code includes: bag of features, 3D MoSIFT, EMoSIFT and SMoSIFT features.
If you use this code, you should cite these two papers.
1) Jun Wan, Qiuqi Ruan, Shuang Deng and Wei Li, "One-shot learning gesture recognition from rgb-d data using bag of features", journal of machine learning research, Vol 14, pp. 2549-2582, 2013.[link: http://jmlr.org/papers/volume14/wan13a/wan13a.pdf]
This paper introduces the 3D EMoSIFT features and also use SOMP to replace VQ for coding descriptors.
2) Jun Wan, Qiuqi Ruan, Wei Li, Gaoyun An and Ruizhen Zhao "3D SMoSIFT: 3D Sparse Motion Scale Invariant Feature Transform for Activity Recognition from RGB-D Videos", Journal of Electronic Imaging, 23(2), 023017, 2014.
This paper introduces the 3D SMoSIFT features. The significant merit is that this new feature is almost real-time. And 3D SMoSIFT can get higher performance than 3D MoSIFT and 3D EMoSIFT features.
3) Jun Wan, Guogong Guo, Stan Z. Li, "Explore Efficient Local Features form RGB-D Data for One-shot Learning Gesture Recognition", under review (submitted to IEEE TPAMI, round 2), 2015
In this paper, Mixed features around sparse keypoints (MFSK) is proposed. The MFSK feaute consist of some popular feature descriptors, such as 3D SMoSIFT, HOG, HOF, MBH. The MFSK feature can outperform all the published approaches on the challenging data of CGD, such as translated, scaled and occludeed subsets.
- Changes to previous version:
Initial Announcement on mloss.org.
No one has posted any comments yet. Perhaps you'd like to be the first?
Leave a comment
You must be logged in to post comments.