XuejingLei

Welcome New MCL Member Abinaya Manimaran

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

I am Abinaya Manimaran, a second year Electrical Engineering graduate student. Before joining USC, I enjoyed working as a Researcher for 3 years at TCS Innovation Labs, India. I was introduced to many Machine Learning algorithms and was always excited to choose the right one to train a desired model. Having joined USC, I am concentrating to learn more in the field of Computer Vision and Machine Learning. I am interested in solving real-world problems that can help mankind be at a better place!

 

2. What is your impression about MCL and USC?

I see MCL as set of hard working and kind students led by experienced Prof Kuo. I like the way how everything is organized, from lab’s website to weekly meetings!

USC is known for very deep course works. Especially when it comes to Electrical Engineering, most of the courses are math intense, which helps in understanding the basics better. Both USC and MCL has a great Alumni network too!

 

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

I would be working on SAAK Transform as a new architecture for deep neural network and its applications in Computer Vision. My plan is to greatly strengthen my skills in this field. I am hoping at the end of my research, I would stand out both personally and professionally.

By |September 16th, 2018|News|Comments Off on Welcome New MCL Member Abinaya Manimaran|
Read More

Welcome New MCL Member Joe Wang

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

My name is Wang, Yun Cheng, but you can simply call me Joe. I like to play basketball, and have been an authentic Lakers’ fan since my junior high. So, it’s exciting for me to be in LA. I came from Taiwan, and currently I am a master student in the electrical engineering department, USC. More specifically, I am in the multimedia and creative technology program, so basically I like to deal with various kinds of multimodal data. I have done some projects on video understanding in my undergraduate, and I hope I can keep on building my strength in multimedia analysis. Recently, I am digging into a novel research topic, network embedding. The research topic is interesting and still has a lot can be explored so I believe I can learn a lot from it. Really looking forward to being in the MCL.

2. What is your impression about MCL and USC?

MCL is a lively lab. Every member in the lab loves what they are doing. They work really hard on what they are enthusiastic about, but at the same time, they know how to enjoy their lives. I think it’s important to know how to strike a good balance between research and the ordinary life for a graduate student. And, members in MCL have just done so well. I enjoy working with the lab members, and we are still friends when we are not in the lab. Discussing with Professor Kuo is also an enjoyable experience, his taste and insight on research topics is so innovative that you can always learn something from him. The campus of USC is gorgeous. But, sometimes it [...]

By |September 9th, 2018|News|Comments Off on Welcome New MCL Member Joe Wang|
Read More

Welcome New MCL Member Mozhdeh Rouhsedaghat

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

My name is Mozhdeh Rouhsedaghat. I received my bachelor’s degree from EE department of Sharif University of Technology. I am currently a Ph.D. student at the Media Communications Lab (MCL) at USC and supervised by professor C. -C. Jay Kuo. My research interests include image processing, computer vision, and deep learning.

 

2. What is your impression about MCL and USC?

USC is a top university and has a beautiful architecture. I really enjoy fresh air in the mornings and walking through amazing buildings on the campus. MCL has a friendly environment and a dynamic atmosphere. Its members are kind and talented and support each other. Professor Kuo is very wise, kind, and caring, and working in this atmosphere is pleasurable for me.

 

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

I am going to improve my research skills in computer vision and I hope I can reach great achievements during my studies.

By |September 2nd, 2018|News|Comments Off on Welcome New MCL Member Mozhdeh Rouhsedaghat|
Read More

Welcome New MCL Member Zhiruo Zhou

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

My name is Zhiruo Zhou, and I just received my bachelor’s degree in Electrical Engineering this July. Now I will start my first year at USC as a PhD student. During my undergraduate training, research projects that I got involved in were mostly about image processing, such as denoising on MRI images and image recognition tasks. Having experiences with both traditional image processing techniques and machine learning, I would like to explore more on their relationship and applications.

2. What is your impression about MCL and USC?

The sunshine here is always warm and the integral architectural style on campus makes me feel comfortable. The MCL is really a big group and has a great network. The pizza lunch and seminar on Friday are fantastic and beneficial. You can always learn from your group members and they are pleased to offer help. It’s enjoyable to know these lovely people, and I’m happy to be one of them!

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

I expect that I could gradually establish a deep understanding of machine learning and be well practiced in doing research. I might also try to refine my career plan by communicating with professors and peers. Also, I want to know more about the MCL family and make friends with them.

By |August 26th, 2018|News|Comments Off on Welcome New MCL Member Zhiruo Zhou|
Read More

Congratulations to Haiqiang Wang for passing PhD defense

Congratulations to Haiqiang Wang for passing his PhD defense on August 21, 2018. His PhD thesis is entitled “A Data-Driven Approach to Compressed Video Quality Assessment Using Just Noticeable Difference”.

Abstract of thesis:
The problem of human-centric compressed video quality assessment (VQA) is studied in the research. In this thesis, he propose a new methodology for compressed video quality measurement and assessment based on the just-noticeable-difference (JND) notion. The process of building a large-scale coded H.264/AVC video quality dataset is described. Then, he measure the JND-based video quality using the satisfied user ratio (SUR) curve and designing an SUR prediction method with video quality degradation feature and masking feature. The method consists of the following steps: 1) partition a video clip into local spatial-temporal segments and evaluate the quality of each segment using the VMAF quality index, 2) aggregate these local VMAF measures to derive a global index, 3) significant segments are selected based on the slope of quality scores between neighboring coded clips, and 4) incorporate the masking effect that reflects the unique characteristics of each video clip. Then use the support vector regression (SVR) to minimize the distance of the SUR curves, and derive the JND point accordingly. At last, propose a JND-based VQA model that takes subject variabilities and content variabilities into account. He build a user model by utilizing user’s capability to discern the quality difference. He study the SUR difference as it varies with user profile as well as content with variable level of difficulty. The proposed model aggregates quality ratings per user group to address inter-group difference.
We are so glad to have him share his PhD experience with us. Here is his sharing.
I would like to thank Professor Kuo for [...]

By |August 21st, 2018|News|Comments Off on Congratulations to Haiqiang Wang for passing PhD defense|
Read More

Welcome New MCL Member – Chenxuan Guo

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

My name is Chenxuan Guo. I am a junior student of Electrical and Computer Engineering department in NCTU (National Chiao Tung University). I come to MCL for an internship in summer. One of the things that I am interested is audio analysis, especially instrumental music separating and converting. Hope one day the accurate realization of audio to score (including audio to midi and midi to score) comes true. The second one is the topic of parallelized computing or distributed operating systems. I am still researching about the faster approximation of optimization of source allocation in fog computing.

2. What is your impression about MCL and USC?

I like the atmosphere of MCL, because Professor Guo’s guidance is very special, more like a father than a teacher. Everyone in the lab has their own strengths and they are good at supporting each other.

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

I hope I can finish my research in graph embedding timely this summer. With help from Pr. Kuo, Bin and Jessica. I must make a push to do my experiment.

 

By |August 19th, 2018|News|Comments Off on Welcome New MCL Member – Chenxuan Guo|
Read More

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|
Read More

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|
Read More

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|
Read More