Xiaofang Wang

I am a final-year Ph.D. student at the Robotics Institute of Carnegie Mellon University. I am fortunate to work with Kris Kitani. Previously, I was a M.S. student at CMU RI, co-advised by Kris Kitani and Martial Hebert. I received my B.S. in Computer Science from Peking University, advised by Ling-Yu Duan.

Email  /  CV  /  Google Scholar  /  Github

Looking for full-time industry jobs starting in Spring 2022.
Welcome to reach out!

profile photo
Research

I am interested in computer vision, deep learning and machine learning. I have worked on neural architecture search, efficient neural networks, network compression, video classification, and image hashing.

Committee-based Wisdom of Committees: An Overlooked Approach To Faster and More Accurate Models
Xiaofang Wang, Dan Kondratyuk, Eric Christiansen, Kris M. Kitani, Yair Alon, Elad Eban
preprint

State-of-the-art efficiency without any architecture tuning
We show that even the most simplistic method for building ensembles or cascades from existing pre-trained networks can attain a significant speedup and higher accuracy over state-of-the-art models.
NANAS Neighborhood-Aware Neural Architecture Search
Xiaofang Wang, Shengcao Cao, Mengtian Li, Kris M. Kitani
British Machine Vision Conference (BMVC), 2021

Finding flat-minima archtectures in the search space
Towards better generalization, we propose a novel neighborhood-aware NAS formulation to identify flat-minima architectures in the search space.
AttentionNAS AttentionNAS: Spatiotemporal Attention Cell Search for Video Classification
Xiaofang Wang, Xuehan Xiong, Maxim Neumann, AJ Piergiovanni, Michael S. Ryoo,
Anelia Angelova, Kris M. Kitani, Wei Hua
European Conference on Computer Vision (ECCV), 2020
[Video-1 minute] [Video] [Slides]

Automatically search for attention cells for video classification
We propose a novel search space for spatiotemporal attention cells and a differentiable search method to learn attention cell designs.
ESNAC Learnable Embedding Space for Efficient Neural Architecture Compression
Shengcao Cao*, Xiaofang Wang*, Kris M. Kitani
International Conference on Learning Representations (ICLR), 2019
* indicates equal contribution.
[Code] [Poster] [Architecture Visualization]

Automatically search for compressed architectures
We propose to learn an embedding space for the architecture domain, based on which we present a compressed architecture search framework using Bayesian optimization.
ErrorCorrection Error Correction Maximization for Deep Image Hashing
Xiang Xu, Xiaofang Wang, Kris M. Kitani
British Machine Vision Conference (BMVC), 2018
DTSH Deep Supervised Hashing with Triplet Labels
Xiaofang Wang, Yi Shi, Kris M. Kitani
Asian Conference on Computer Vision (ACCV), 2016
Oral Presentation, (5.6% acceptance rate)
[Code]
HCQ Hamming Compatible Quantization for Hashing
Zhe Wang, Ling-Yu Duan, Jie Lin, Xiaofang Wang, Tiejun Huang, Wen Gao
International Joint Conference on Artificial Intelligence (IJCAI), 2015
Industry Experience
Google Research Google Perception
Research Intern
May 2020 - August 2020
Google Cloud Google Cloud AI
Research Intern
May 2019 - August 2019
Service

Journal Reviewer: IJCV, TIP, ACM Computing Surveys

Conference Reviewer: CVPR, ICCV, ECCV, NeurIPS, ICML, ICLR


Website design from Jon Barron