I am a master student in National Tsing Hua University (NTHU) from Taiwan, and I am supervised by Prof. Chun-Yi Lee. My research area including autonomous car, image segmentation, and robotics. Curently, I am looking for internship opportunity!

See my CV for more detail.


Research Assistant

2017 - present

My research focuses on computer vision and robotics. Supervised by Prof. Chun-Yi Lee, I worked on several research projects which published in CVPR'18, IJCAI'18, and ECCV'18.

Teaching Assistant in DL workshop

2017 - present
NVIDIA Deep Learning Institute

I am responsible for guiding students in NVIDIA workshops, including the tutorials of CUDA parallel programming, object detecion, and image segmentation.

Summer Intern at CyberLink Corp.

Jun. 2016 - Sep. 2016
Power Director Team - Research and devolopment engineer

During the internship, I implemented new image/video effects and solved critical issues in Power Director released versions. Also, I improved algorithms to be released in future version.


Dynamic Video Segmentation Network
Y.-S Xu, H.-K Yang*, T.-J Fu*, and C.-Y Lee
Published in CVPR 2018
[Paper] [Project] [Code]
Virtual-to-Real: Learning to Control in Visual Semantic Segmentation
Z.-W Hong, H.-K Yang*, Y.-M Chen*, S.-Y Su*, T.-Y Shann*, and C.-Y Lee
Published in IJCAI 2018
[Paper] [Project] [Code]
Visual Relationship Prediction via Label Clustering and Incorporation of Depth Information
H.-K Yang, A.-C Cheng, K.-W Ho, T.-J Fu, and C.-Y Lee
Publisehed in ECCV 2018 (workshop)
[Paper] [Project] [Code]
Embedding Cluster for Accelerating Semantic Video Segmentation
H.-K Yang*, T.-J Fu*, K.-W Ho, P.-H Chiang and C.-Y Lee
Under review


Semantic Segmentation with Deep Learning - Implement several state-of-art semantic segmentation network based on Tensorflow framework, including ICNet, PSPNet, Deeplab, and FCN. These model trained on the Cityscapes and ADE20k dataset. The cityscapes dataset focuses on semantic understanding of urban street scenes, while ADE20k dataset focuses on both indoor scenes and outdoor scenes.
Transforming Graffitis to Realistic Sketches using GANs - The goal of this project is to convert car gra ffiti pictures into a more complicated, realistic drawing. We used CycleGAN as our baseline. What we have achieved in this project is having aids from additional image encoded with input, to assist the model find the mapping more easily to the desired output.


2017    Scholarship for EECS Excellent Students (for Top 5%)
2017    Computer Science Senior Project Contest - 1st prize
2018    The Cutting Edge of Deep Learning Contest - 3rd prize
2018    College of Eletrical Engineering and Computer Science Contest - 1st prize
2018    NVIDIA Jetson Developer Challege Contest in GTC2018 - 1st prize (1/114) 2018    1st People In Context (PIC) Challenge in ECCV2018 - 2nd prize

Extracurricular Activities