Zhuo Su

I am an AI Research Scientist at Huawei Helsinki R&D Center in Finland. I am also a PostDoc (part-time) at Center for Machine Vision and Signal Analysis (CMVS), University of Oulu, in Generative AI.

My doctoral thesis defended in October: LBP Inspired Efficient Deep Convolutional Neural Networks for Visual Representation Learning. Involved topics are:

Local binary pattern (LBP) and LBP inspired CNN modules for efficient visual representation learning.
Binary neural networks on regular (images) and irregular data (point clouds).
Dynamic networks and network pruning.

My interests: Machine learning, Deep Learning, Computer vision, Generative AI, Multimodal Learning, Math, Coding.

Email / CV / Google Scholar / Github / LinkedIn / Blogs

Latest News

[9.2024] I become a AI Research Scientist at Huawei Helsinki Research Center since September 30. Part-time work in the University of Oulu as a PostDoc.
[7.2024] One paper got major revision from IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). Title: Rapid Salient Object Detection with Difference Convolutional Neural Networks.
[1.2024] One paper got accepted from IEEE Transactions on Neural Networks and Learning Systems (TNNLS). Title: Boosting Convolutional Neural Networks with Middle Spectrum Grouped Convolution.
[11.2023] I am now a PostDoc at Center for Machine Vision and Signal Analysis (CMVS), University of Oulu.
[10.2023] I defended my doctoral thesis titled LBP Inspired Efficient Deep Convolutional Neural Networks for Visual Representation Learning at University of Oulu. Supervisor: Dr. Li Liu, University of Oulu; Oppoenent: Professor Karen Eguiazarian, Tampere University; Custos: Professor Matti Pietikäinen, University of Oulu.
[7.2023] One paper got accepted from IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). Title: Lightweight Pixel Difference Networks for Efficient Visual Representation Learning.

Experience and Education

[11.2018 - Present] PostDoc researcher in Computer Science at University of Oulu.
[10.2018 - 10.2023] Doctoral researcher in Computer Science at University of Oulu, supervised by Li Liu.
[09.2022 - 03.2023] Computer vision intern at Intel Labs, Germany, working with Matthias Müller.
[10.2021 - 03.2022] Visiting researcher in AMLab, University of Amsterdam, supervised by Max Welling.
[10.2018 - precent] PhD candidate of Computer vision in University of Oulu.
[05.2018 - 07.2018] Software engineer in Samsung R&D Institute, China-Beijing.
[01.2018 - 04.2018] Computer vision intern in Aihujing.com.
[09.2015 - 03.2018] Master student in Beihang University (thesis: Salient object detection for single images, supervisor: Hong Zheng).
[09.2011 - 06.2015] Bachelor student in Beihang University (area: Pattern recognition).

Publications (First author)

	Boosting Convolutional Neural Networks with Middle Spectrum Grouped Convolution Zhuo Su, Jiehua Zhang, Tianpeng Liu, Zhen Liu, Shuanghui Zhang, Matti Pietikäinen, Li Liu TNNLS (IEEE Transactions on Neural Networks and Learning Systems), 2024 Efficient neural network backbones, image Classification, object detection.
	Lightweight Pixel Difference Networks for Efficient Visual Representation Learning Zhuo Su, Jiehua Zhang, Longguang Wang, Hua Zhang, Zhen Liu, Matti Pietikäinen, Li Liu TPAMI (IEEE Transactions on Pattern Analysis and Machine Intelligence), 2023 Efficient networks for edge detection, object classification, and facial recognition.
	From Local Binary Patterns to Pixel Difference Networks for Efficient Visual Representation Learning Zhuo Su, Matti Pietikäinen, Li Liu SCIA (Scandinavian Conference on Image Analysis), 2023 arxiv A brief survey on Local Binary Pattern (LBP) inspired Deep Learning networks.
	SVNet: Where SO(3) Equivariance Meets Binarization on Point Cloud Representation Zhuo Su, Max Welling, Matti Pietikäinen, Li Liu 3DV (3D Vision), 2022 arxiv / code SVNet is a general framework to construct 3D learning architectures with SO(3) equivariance and network binarization, with a novel dual use of scalar and vector features.
	Pixel Difference Networks for Efficient Edge Detection Zhuo Su, Wenzhe Liu, Zitong Yu, Dewen Hu, Qing Liao, Qi Tian, Matti Pietikäinen, Li Liu ICCV (oral presentation), 2021 arxiv / code PDC embeds traditional Local Binary Pattern (LBP) to convolutional operators to enable CNNs probe rich image gradient information.
	Dynamic Group Convolution for Accelerating Convolutional Neural Networks Zhuo Su, Linpu Fang, Wenxiong Kang, Matti Pietikäinen, Li Liu ECCV (spotlight presentation), 2020 arxiv / code Introducing dynamic execution in grouped convolution to enhance existing CNN backbones.
	BIRD: Learning Binary and Illumination Robust Descriptor for Face Recognition Zhuo Su, Matti Pietikäinen, Li Liu BMVC, 2019 pdf / code BIRD (Binary and Illumination Robust Descriptor) nicely balances the three criteria: distinctiveness, robustness, and computationally inexpensive cost for face representation.

Publications (Co-author)

	Highly Efficient and Unsupervised Framework for Moving Object Detection in Satellite Videos Chao Xiao, Wei An, Yifan Zhang, Zhuo Su, Miao Li, Weidong Sheng, Matti Pietikäinen, Li Liu TPAMI (IEEE Transactions on Pattern Analysis and Machine Intelligence), 2024 An unsupervised framework with sparse convolutional neural networks for highly efficient moving object detection.
	Enhancing Information Maximization With Distance-Aware Contrastive Learning for Source-Free Cross-Domain Few-Shot Learning Huali Xu, Li Liu, Shuaifeng Zhi, Shaojing Fu, Zhuo Su, Ming-Ming Cheng, Yongxiang Liu TIP (IEEE Transactions on Image Processing), 2024 Source-free cross-domain few-shot learning with transductive learning.
	Dynamic Binary Neural Network by learning channel-wise thresholds Jiehua Zhang, Zhuo Su, Yanghe Feng, Xin Lu, Matti Pietikäinen, Li Liu ICASSP, 2022 arxiv Using dynamic thresholds to boost binary neural networks.
	Median Pixel Difference Convolutional Network for Robust Face Recognition Jiehua Zhang, Zhuo Su, Li Liu BMVC, 2021 arxiv A novel combination between MRELBP (a traditional LBP variant) and convolution to build robust CNNs for face recognition in noisy conditions.
	Deep ladder reconstruction-classification network for unsupervised domain adaptation Wanxia Deng, Zhuo Su, Qiang Qiu, Lingjun Zhao, Gangyao Kuang, Matti Pietikäinen, Huaxin Xiao, Li Liu Pattern Recognition Letters, 2021 Deep Ladder Reconstruction-Classification Network (DLaReC) is designed to learn cross-domain shared contents by suppressing domain-specific variations.
	Searching central difference convolutional networks for face anti-spoofing Zitong Yu, Chenxu Zhao, Zezheng Wang, Yunxiao Qin, Zhuo Su, Xiaobai Li, Feng Zhou, Guoying Zhao CVPR, 2020 arxiv / code Central Difference Convolution (CDC) is able to capture intrinsic detailed patterns via aggregating both intensity and gradient information for face anti-spoofing.

Activities

Teaching assistant in University of Oulu:

[Autumn 2022] TA of Deep Learning course (521153S-3003, Li Liu), along with Jiehua Zhang, Shuzhou Sun, and Huali Xu.
[Autumn 2021] TA of Deep Learning course (521153S:3, Li Liu), along with Jiehua Zhang, Huali Xu, and Lam Huynh.
[Autumn 2020] TA of Deep Learning course (521153S:2, Li Liu), along with Lam Huynh, Mohammad Tavakolian, and Yawen Cui.
[Autumn 2019] TA of Deep Learning course (521153S, Li Liu and Jie Chen), along with Lam Huynh.
[Spring 2018] TA of Deep Learning course (521149S, Li Liu and Jie Chen), along with Lam Huynh.

Journals: IEEE Transactions on Pattern Analysis and Machine Intelligence, IEEE Transactions on Image Processing, IEEE Transactions on Circuits and Systems for Video Technology, Neurocomputing, Pattern Recognition Letters.
Conferences: CVPR, ECCV, AAAI, ICME, ICCV, ACCV, ICASSP.

Template from Jon Barron. Thanks!