Tsung-Yu Lin

Research Assistant
College of Information and Computer Sciences
University of Massachusetts Amherst
tsungyulin at cs.umass.edu


I am a sixth year Computer Science PhD student at University of Massachusetts Amherst working at Computer Vision Lab led by Prof. Erik Learned-Miller and Prof. Subhransu Maji. In general I am working in Computer Vision and Machine Learning. Specifically my research interests are in designing machine learning models to understand visual contents from images or videos. This includes learning semantic visual representations, modeling objects geometry and appearance, and designing generalizable models to the data in real world.

During my PhD study, I have been working as intern with Prof. Tamara Berg and and Prof. Alex Berg at Facebook AML. I had been working as intern with Prof. R. Manmatha and Prof. Deva Ranmanan at Amazon AI. Before I started my PhD study, I had been working at Academia Sinica with Dr. Tyng-Luh Liu. I completed my MS degree at National Tsing Hua University advised by Prof. Shang-Hong Lai. Check out my CV and Google scholar profile.





Second-order Democratic Aggregation
Dark Ecology: tracking bird migration with computer vision and deep learning techniques
Visualizing and Understanding Deep Texture Representations
Bilinear CNN for Fine-Grained Classification
People localization in a camera network
Shape Prior Non-Rigid SfM for deformable Surfaces
Multi-camera and multi-target surveillance tracking 3D Face Model Deformation


MistNet: Measuring Historical Bird Migration in the US Using Archived Weather Radar Data and Convolutional Neural Networks

To appear in Methods in Ecology and Evolution, 2019    [project]

Second-order Democratic Aggregation

ECCV 2018    [project] [pdf] [arXiv]

Improved Bilinear Pooling with CNNs

BMVC 2017 (Oral)    [project] [pdf] [arXiv]

Bilinear Convolutional Neural Networks for Fine-grained Visual Recognition

PAMI 2017   [project] [pdf]

Visualizing and Understanding Deep Texture Representations

CVPR 2016    [project] [pdf] [arXiv]

One-to-many face recognition with Bilinear CNNs

WACV 2016   [pdf]

Implicit Sparse Code Hashing

arXiv:1512.00130 , 2015    [arXiv]

Bilinear CNN Models for Fine-grained Visual Recognition

ICCV 2015  (Oral, Acceptance rate: 3.3%)    [project] [pdf] [pdf-supp] [arXiv] [slides] [poster] [bibtex]

Efficient binary codes for extremely high-dimensional data

ICIP 2014  [pdf]

People localization in a camera network combining background subtraction and scene-aware human detection

MMM 2011  [pdf]