Starting from applications related to hand detection and posture analysis, I love to solve real-world problems with economic, explainable and reproducible solutions.
Parametrix.ai
What is the difference between AIs and living beings, if they have similar intellectural levels, or AIs are even smarter?
Tencent TEG Group x 腾讯微视
FeatherCNN is a lightweight and fast inference framework for Arm Neon CPUs.
Tencent TEG Group x QQ Mobile
Real-time skeleton tracking and action recognition on iOS and Android smart phones shipped since 2014.
Tencent TEG Group x QQ Mobile
Image filters for live video style transfer on iOS and Android smart phones.
Tencent TEG Group
A filter that changes your photo background with Chinese New Year greetings.
Tencent TEG Group x PTu
An interesting content-aware photo filter.
Tencent TEG Group x PTu
Neural style photo filter is possible in Milliseconds!
Tencent TEG Group
I develop algorithms for distributed training of acoustic models.
Xiaolong Zhu, Wei Liu, Xuhui Jia and Kwan-Yee K. Wong
WACV 2016
Dataset is available upon request.
Now the end-to-end hand detection is possible via multi-task learning for CNN.
Xiaolong Zhu, Xuhui Jia and Kwan-Yee K. Wong
ACCV 2014
CVIU 2015
Dataset is available upon request.
Pixel level hand detection using structured forests. It is very efficient and robust.
Internship in Lenovo IVC Lab.
Xiaolong Zhu, Ruoxin Sang, Xuhui Jia and Kwan-Yee K. Wong
IVCNZ 2013
Dataset is available upon request.
Provide possibility to sketch a hand shape recognizer using a customized Boltzmann machine.
Internship in MSRA HCI Group.
Xiaolong Zhu and Kwan-Yee K. Wong
ICPR 2012
Dataset is available upon request.
Single frame hand posture recognition by kernel descriptor for efficient SVM prediction.
Zhihu Chen, Kwan-Yee K. Wong, Yasuyuki Matsushita, Xiaolong Zhu and Miaomiao Liu
ICCV 2011
IJCV 2013
Obtain the depth of the image by refraction. Easy setup, promising results!
Front-end development for an LBS application.
Xiaolong Zhu, Huijing Zhao, Yiming Liu, Yipu Zhao and Hongbin Zha
IROS 2010
ICRA 2010
Semantic labelling for range data. Good segmentation is the key.