MCL Research on Efficient Text Classification

A novel text data dimension reduction technique, called the tree-structured multi-linear principal component analysis (TMPCA), is proposed in this work. Being different from traditional text dimension reduction methods that deal with the word-level representation, the TMPCA technique reduces the dimension of input sequences and sentences to simplify the following text classification tasks. It is shown mathematically and experimentally that the TMPCA tool demands much lower complexity (and, hence, less computing power) than the ordinary principal component analysis (PCA).  Furthermore, it is demonstrated by experimental results that the support vector machine (SVM) method applied to the TMPCA-processed data achieves commensurable or better performance than the state-of-the-art recurrent neural network (RNN) approach.


by Yuanhang Su

By |March 26th, 2018|News, Research|Comments Off on MCL Research on Efficient Text Classification|

MCL Research Presented at WACV 2018

MCL member, Ye Wang presented Qin Huang’s paper at Winter Conference on Applications of Computer Vision (WACV) 2018, Lake Tahoe, NV/CA

The title of Qin’s paper is “Unsupervised Clustering Guided Semantic Segmentation”, with Chunyang Xia, Siyang Li, Ye Wang, Yuhang Song and C.-C. Jay Kuo as the co-authors. Here is a brief summary:

“With the development of Fully Convolutional Neural Network (FCN), there have been progressive advances in the field of semantic segmentation in recent years. The FCN-based solutions are able to summarize features across training images and generate matching templates for the desired object classes, yet they overlook intra-class difference (ICD) among multiple instances in the same class. In this work, we present a novel fine-to-coarse learning (FCL) procedure, which first guides the network with designed ‘finer’ sub-class labels, whose decisions are mapped to the original ‘coarse’ object category through end-to-end learning. A sub-class labeling strategy is designed with unsupervised clustering upon deep convolutional features, and the proposed FCL procedure enables a balance between the fine-scale (i.e. sub-class) and the coarse-scale (i.e. class) knowledge. We conduct extensive experiments on several popular datasets, including PASCAL VOC, Context, Person-Part and NYUDepth-v2 to demonstrate the advantage of learning finer sub-classes and the potential to guide the learning of deep networks with unsupervised clustering.”

Congratulations to Qin Huang for his successful presentation at WACV!

By |March 18th, 2018|News|Comments Off on MCL Research Presented at WACV 2018|

Welcome New MCL Member – Alex Ding

We are so happy to welcome a new undergraduate member of MCL, Alex Ding! Here is an interview with Alex:

1. Could you briefly introduce yourself and your research interests?

My name is Yulong (Alex) Ding and I am a third-year undergraduate student at USC with a major in Electrical Engineering. I plan to further pursue studies in the field of EE at graduate level after I complete my bachelor’s degree and I find the application of machine learning and CNNs in data and image processing at MCL as an exciting area for further exploration in the future. At the same time, I would like to gain insight into the process of conducting research in the academic field through valuable experiences at MCL.

2. What is your impression about MCL and USC?

Having attended the weekly seminars held by Professor Kuo and the MCL community, I felt inspired by the various topics of research presented as well as the passion shared by members of MCL. The potential opportunities surrounding data processing and communication that one may explore at MCL are fascinating. Furthermore, everyone in the MCL community, including both current students and alumni, are very supportive, building upon the well-known Trojan network connecting across all career paths.

3. What is your future expectation and plan in MCL?

My current goal at MCL is to acquire knowledge in data processing and machine learning, as well as gain experience in the process of conducting research through involvement in the lab. The valuable experiences at MCL will serve as important preparation for my future pursuit in studies and research at graduate level, and ultimately for potential career opportunities in the relevant field.

By |March 11th, 2018|News|Comments Off on Welcome New MCL Member – Alex Ding|

Welcome New MCL Member – Charan Shettyhalli Guruswamy

We are so happy to welcome a new Master’s member of MCL, Charan Shettyhalli Guruswamy. Here is an interview with Charan :
1. Could you briefly introduce yourself and your research interests?
I am Charan, a new grad student who joined EE dept in spring 2018. I love to work hard and think more. I got attracted to image processing and machine learning during my undergrad. Starting then I have pursued computer vision and the math behind it. My previous research included face & object detection and recognition, biometrics, NLP, optimization, ML and deep learning. I have a passion to play with numbers, and I see images as just matrices. I am not a big fan of neural networks, I like convolutions more than deep learning. I have interest in designing new transforms for data, I hope to research on scene understanding and video completion.

2. What is your impression about MCL and USC?
MCL is a group filled with image processing passionates. Every one of them has their own beautiful perspectives in analyzing the images, and their unique works enthrall me. There is a lot of math, analysis, and visualization here, that’s what I like in MCL. Weekly sessions by doctoral students help in understanding the new topics in the IP field, thus triggering motivation in all. The rich MCL alumni and community guiding the current MCL candidates is ideal for career growth. MCL helps in breaking our shell and surpass our limits.

3. What is your future expectation and plan in MCL?
My goal is to design a transforms which can extract and select features from any data. I want to learn how to model this transforms. I want to join the peer community of [...]

By |March 4th, 2018|News|Comments Off on Welcome New MCL Member – Charan Shettyhalli Guruswamy|

MCL Research on Saak Transform

It is well known that CNNs-based methods have weaknesses in terms of efficiency, scalability, and robustness. CNNs-based methods require large computational efforts and are not scalable to the change of object class numbers and the dataset size. Furthermore, these CNN models are not robust to small perturbations due to their excess dependence on the end-to-end optimization methodology. The Saak (Subspace approximation with augmented kernels) transform [1] is proposed to provide the possible solution to overcome these shortcomings.

The Saak transform consists of two new ingredients on top of traditional CNNs. They are: subspace approximation and kernel augmentation. The Saak transform allows both forward and inverse transforms so that it can be used for image analysis as well as synthesis (or generation). One can derive a family of joint spatial-spectral representations between two extremes – the full spatial-domain representation and the full spectral-domain representation using multi-stage Saak transforms. Being different with CNNs, all transform kernels in multi-stage Saak transforms are computed by one-pass feedforward process. Neither data labels nor backpropagation is needed for kernel computation.

Currently, we have successfully developed Saak transform approaches [2] to solve the handwritten digits recognition problem. This new approach has several advantages such as higher efficiency than the lossless Saak transform, scalability against the variation of training data size and object class numbers and robustness against noisy images. In the near future, we would like to apply the Saak transform approach to the general object classification problem with more challenging datasets such as CIFAR-10, CIFAR-100, and ImageNet.



[1]  C-C Jay Kuo and Yueru Chen, “On Data-driven Saak Transform,” arXiv preprint arXiv:1710.04176, 2017.

[2] Chen, Y., Xu, Z., Cai, S., Lang, Y., & Kuo, C. C. J. (2017). A Saak Transform Approach to Efficient, Scalable [...]

By |February 25th, 2018|News|Comments Off on MCL Research on Saak Transform|

MCL Research on Image Splicing Localization

With the advent of Web 2.0 and ubiquitous adoption of low-cost and high-resolution digital cameras, users upload and share images on a daily basis. This trend of public image distribution and access to user-friendly editing software such as Photoshop and GIMP has made image forgery a serious issue. Splicing is one of the most common types of image forgery. It manipulates images by copying a region from one image (i.e., the donor image) and pasting it onto another image (i.e., the host image). Forgers often use splicing to give a false impression that there is an additional object present in the image, or to remove an object from the image. Image splicing can be potentially used in generating false propaganda for political purposes. For example, during the 2004 U.S. presidential election campaign, an image that showed John Kerry and Jane Fonda speaking together at an anti-Vietnam war protest was released and circulated. It was discovered later that this was a spliced image, and was created for political purposes. The spliced image and the two original authentic images that were used to create the spliced image can be seen above.


Many of the current splicing detection algorithms only deduce whether a given image has been spliced and do not attempt to localize the spliced area. Relatively few algorithms attempt to tackle the splicing localization problem, which refers to the problem of determining which pixels in an image have been manipulated as a result of a splicing operation.


Ronald Salloum and Professor Jay Kuo are currently working on an image splicing localization research project. They are exploring the use of deep learning and data-driven techniques to develop an effective solution to the problem of image splicing localization. They [...]

By |February 19th, 2018|News, Research|Comments Off on MCL Research on Image Splicing Localization|
  • Permalink Gallery

    Congratulations to MCL Alumnus Dr. Eddy Wu for Joining Megvii/Face++

Congratulations to MCL Alumnus Dr. Eddy Wu for Joining Megvii/Face++

We would like to say congratulations to Dr. Eddy Wu for graduating from USC/MCL. He is going to join a start-up company in Seattle in 2018 February. He prepared a message about his job hunting experience to share with all MCL members, Alums and friends. We wish him the very best in his career at Magvii Research.

After graduating from MCL, I’ve decided to join Megvii Research USA as a senior research scientist. Megvii is a fast-growing unicorn startup company based in Beijing, China focusing on AI technologies. It is especially famous for its face recognition technology (known as Face++). Megvii research USA is a new branch located at Redmond WA, and is directed by Dr. Jue Wang ( The whole Megvii research is led by Dr. Jian Sun (

Joining a startup company is always risky, but I have a strong desire becoming part of Megvii for several reasons. First, I like fast-paced environments, and I enjoy the challenges of bringing latest computer vision and deep learning technologies to real-world applications. Secondly, Megvii has shown great successes in certain AI fields with their talented teams, and signs have shown that China may have potential to lead in AI development in the foreseeable future. Another reason is that my new role in Megvii also has a great match with my researches in MCL as well as my 6-year industrial experiences prior to my PhD study.

Job hunting is an exhausted process. I applied for, and interviewed with different types companies including big name ones, industrial research labs, and startup companies in different stages. I only targeted computer vision or deep learning roles, with different types (engineer/scientist) and levels. Although I got interview invitations from many companies, eventually [...]

By |February 14th, 2018|News|Comments Off on Congratulations to MCL Alumnus Dr. Eddy Wu for Joining Megvii/Face++|

MCL Research on Unsupervised Video Segmentation

We propose a method for unsupervised video object segmentation by transferring the knowledge encapsulated in image-based instance embedding networks. The instance embedding network produces an embedding vector for each pixel that enables identifying all pixels belonging to the same object. Though trained on static images, the instance embeddings are stable over consecutive video frames, which allow us to link objects together over time. Thus, we adapt the instance networks trained on static images to video object segmentation and incorporate the embeddings with objectness and optical flow features, without model retraining or online fine-tuning. The proposed method outperforms state-of-the-art unsupervised segmentation methods in the DAVIS dataset and the FBMS dataset.

The main contributions include

– A new strategy for adapting instance segmentation models trained on static images to videos. Notably, this strategy performs well on video datasets without requiring any video object segmentation annotations.

– Proposal of novel criteria for selecting a foreground object without supervision, based on semantic score and motion features over a track.

– Insights into the stability of instance segmentation embeddings over time.


By Siyang Li

By |February 12th, 2018|News, Research|Comments Off on MCL Research on Unsupervised Video Segmentation|

Welcome New MCL Member – Sibo Song

We are so happy to welcome a new Ph.D. member of MCL, Sibo Song. Here is an interview with  Sibo :
1. Could you briefly introduce yourself and your research interests?

My name is Sibo Song. I am currently a Ph.D. student at Singapore University of Technology and Design (SUTD) and I received my Bachelor’s degree from Zhejiang University(ZJU) in Automation. My research interests include activity recognition, multi-modal data analysis, and deep learning.

2. What is your impression about MCL and USC?

I know MCL from Prof. Kuo when he visited Singapore and presented his recent works on Saak transform. Through the talking, I knew that MCL is one big family of many intelligent people who are willing to offer their help and enthusiastic about doing research. I just can’t wait to join MCL and meet them all. 

3. What is your future expectation and plan in MCL?

During the exchange, I plan to receive research training and investigate Saak transform on various applications such as generative model and adversarial attacks in image and video processing. Also, I hope that I can make more friends here in MCL and learn together with them.

By |February 4th, 2018|News|Comments Off on Welcome New MCL Member – Sibo Song|

Welcome New MCL Member – Tianxiao Zhang

We are so happy to welcome a new Master’s member of MCL, Tianxiao Zhang. Here is an interview with Tianxiao:

1. Could you briefly introduce yourself and your research interests?

I am Tianxiao Zhang, and I just graduated from USC in December 2017 and received my M.S. degree, and I joined the Media Communication Lab in January 2018, guided by professor C.-C. Jay Kuo. I love mathematics and programming and my research interests are deep learning and image processing. I think the future belongs to deep learning and deep learning belongs to image understanding. By the way, I love music and I am also a professional pianist. Hope I can combine these two different areas and do a good job in deep learning in the future.
2. What is your impression about MCL and USC?

MCL is like a great family and we have meetings for all MCL members every week. In the meeting, I could meet with other members and make friends with them. I really love the atmosphere in MCL, happy and friendly. What’s more, professor Kuo, the director of MCL, is not only an excellent professor but also a hard-working person. I have never seen any professor starting their work at 6:30 am in the morning until I meet professor Kuo. So I can learn a lot from professor Kuo not only the knowledge but also how to work hard. 
3. What is your future expectation and plan in MCL?

I will cooperate with other MCL members and follow the instructions of professor Kuo, and contribute as much as I can to the research in MCL, especially in deep learning.

By |January 27th, 2018|News|Comments Off on Welcome New MCL Member – Tianxiao Zhang|