Recent & Highlighted News

More Posts

On June 6th, 2019, C3SR members were invited to present the center’s work on MLModelScope at the MLPerf community’s meeting in Santa Clara, CA. At the meeting, Dr. Jinjun Xiong on behalf of the team presented “Benchmarking and Understanding ML Inference,” a talk drawing on research results from MLModelScope. Among the MLModelScope’s team members, Ms. Cheng Li also attended the meeting in person, while Mr. Abdul Dakkak attended the meeting online.

CONTINUE READING

For the second year in a row, our C3SR teams are victorious in a competition challenging experts worldwide to design low-power embedded systems for Internet-of-Things (IoT) applications. At the 2019 Design Automation Conference (DAC) System Design Contest, our teams won Double Championships for both the GPU and the FPGA tracks of the contest. The contest’s objective is to create algorithms that can accurately detect and locate objects from images taken by aerial drones.

CONTINUE READING

C3SR faculty, Prof. Lav Varshney, was invited to give a talk at the opening ceremony of the 2019 National Science Olympiad competition. There were more than 2,000 top middle and high school students from around the country.

Here is the news story.

Prof. Varshney’s talk was well received by the thousands of enthusiastic students.

CONTINUE READING

On April 17, 2019, C3SR members Ms. Cheng Li, Mr. Abdul Dakkak, Prof. Jinjun Xiong and Prof. Wen-mei Hwu were invited to give a tutorial on the center’s work on MLModelScope at the 2019 ASPLOS conference in Providence, RI.

The tutorial was well received by the attendees. Details about this tutorial can be found here.

CONTINUE READING

One of the joint research work between the C3SR team and the IBM Research, titled “Evaluating Characteristics of CUDA Communication Primitives on High-Bandwidth Interconnects,” won the Best Research Paper Award at the 10th ACM/SPEC International Conference on Performance Engineering (ICPE).

The winning team members were Carl Pearson, Abdul Dakkak, Sarah Hashash, Cheng Li, I-Hsin Chung, Jinjun Xiong, and Wen-mei Hwu.

Congratulations!

CONTINUE READING

Publications

More Publications

ArXiv 2018, 2018

2018 International Conference On Computer Aided Design, 2018

The 51st Annual IEEE/ACM International Symposium on Microarchitecture, 2018

Recent region-based object detectors are usually built with separate classification and localization branches on top of shared feature extraction networks. In this paper, we analyze failure cases of state-of-the-art detectors and observe that most hard false positives result from classification instead of localization. We conjecture that: (1) Shared feature representation is not optimal due to the mismatched goals of feature learning for classification and localization; (2) multi-task learning helps, yet optimization of the multi-task loss may result in sub-optimal for individual tasks; (3) large receptive field for different scales leads to redundant context information for small objects.We demonstrate the potential of detector classification power by a simple, effective, and widely-applicable Decoupled Classification Refinement (DCR) network. DCR samples hard false positives from the base classifier in Faster RCNN and trains a RCNN-styled strong classifier. Experiments show new state-of-the-art results on PASCAL VOC and COCO without any bells and whistles.
European Conference on Computer Vision, 2018

European Conference on Computer Vision, 2018

European Conference on Computer Vision, 2018

28th International Conference on Field-Programmable Logic and Applications, 2018

Frontiers in ICT, 2018

The 56th Annual Meeting of the Association for Computational Linguistics, 2018

The 56th Annual Meeting of the Association for Computational Linguistics, 2018

Invited Talks

Cloud Tools and Libraries for Exploriting Heterogeneous Cognitive Computing Systems
Wed, Sep 27, 2017
Architecture and Software for Emerging Low-Power Systems
Mon, Jul 24, 2017
Crowdsensing, Crowdsourcing, and Creativity
Thu, Jun 1, 2017
Cognitive Computing on Heterogeneous Hardware Systems for the AI Revolution
Sat, Jul 8, 2017
Innovative Applications and Technology Pivots – A Perfect Storm in Computing
Sat, Feb 11, 2017
Best Research Paper Award, "Evaluating Characteristics of CUDA Communication Primitives on High-Bandwidth Interconnects,"
Thu, Apr 11, 2019
Best Paper Award, "DNNBuilder: an Automated Tool for Building High-Performance DNN Hardware Accelerators for FPGAs,"
Mon, Nov 5, 2018
Third Place, 2018 ICPC Mid-Central USA Programming Contest, Team: Haoming Lu, Yunqi Li, Linyi Li
Sat, Nov 3, 2018
Best Paper Award Nomination, "Application-Transparent Near-Memory Processing Architecture with Memory Channel Network"
Sat, Oct 20, 2018
The C3SR team was one of the Finalists for the 2018 Champions at the 2018 IEEE GraphChallenge and will present their work at the 2018 IEEE HPEC (High Performance Extreme Computing) conference
Tue, Sep 25, 2018
The C3SR team won the Student Innovation Award at the 2018 IEEE GraphChallenge and will present their work at the 2018 IEEE HPEC (High Performance Extreme Computing) conference
Tue, Sep 25, 2018
The C3SR team won the 3rd place for the DAC 2018 System Design Competition
Mon, Jun 25, 2018
The C3SR team won the 3rd place for the Nvidia AI City Challenge for traffic flow estimation at 2018 CVPR workshop on NVIDIA AI CITY CHALLENGE
Mon, Jun 18, 2018