Jung-Woo Chang

Jung-Woo Chang

Auto-generated profile, based on publicly available information

Jung-Woo Chang

University of California San Diego

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
20 December 2018 in IEEE Transactions on Circuits and Systems for Video Technology
Verified
Journal Article
Closed access
JCJung-Woo Chang
Jung-Woo Chang
KKKeon-Woo Kang
Keon-Woo Kang
SKSuk-Ju Kang
Suk-Ju Kang
Convolutional neural networks (CNNs) demonstrate excellent performance in various computer vision applications. In recent years, FPGA-based CNN accelerators have been proposed for optimizing performance and power efficiency. Most accelerators are designed for object detection and recognition algorithms that are performed on low-resolution (LR) images. However, real-time image super-resolution (SR) cannot be implemented on a typical accelerator because of the long execution cycles required to generate high-resolution (HR) images, such as those used in ultra-high-definition (UHD) systems. In this paper, we propose a novel CNN accelerator with efficient parallelization methods for SR applications. First, we propose a new methodology for optimizing the deconvolutional neural networks (DCNNs) used for increasing feature maps. Secondly, we propose a novel method to optimize CNN dataflow so that the SR algorithm can be driven at low power in display applications. Finally, we quantize and compress a DCNN-based SR algorithm into an optimal model for efficient inference using on-chip memory. We present an energyefficient architecture for SR and validate our architecture on a mobile panel with quad-high-definition (QHD) resolution. Our experimental results show that, with the same hardware resources, the proposed DCNN accelerator achieves a throughput up to 108 times greater than that of a conventional DCNN accelerator. In addition, our SR system achieves an energy efficiency of 144.9 GOPS/W, 293.0 GOPS/W, and 500.2 GOPS/W at SR scale factors of 2, 3, and 4, respectively. Furthermore, we demonstrate that our system can restore HR images to a high quality while greatly reducing the data bit-width and the number of parameters compared to conventional SR algorithms.

9 December 2022 in ACM Journal on Emerging Technologies in Computing Systems
Verified
Journal Article
Open access
MJMojan Javaheripi
Mojan Javaheripi
JCJung-Woo Chang
Jung-Woo Chang
FKFarinaz Koushanfar
Farinaz Koushanfar
We propose AccHashtag , the first framework for high-accuracy detection of fault-injection attacks on Deep Neural Networks (DNNs) with provable bounds on detection performance. Recent literature in fault-injection attacks shows the severe DNN accuracy degradation caused by bit flips. In this scenario, the attacker changes a few DNN weight bits during execution by injecting faults to the dynamic random-access memory (DRAM). To detect bit flips, AccHashtag extracts a unique signature from the benign DNN prior to deployment. The signature is used to validate the model’s integrity and verify the inference output on the fly. We propose a novel sensitivity analysis that identifies the most vulnerable DNN layers to the fault-injection attack. The DNN signature is constructed by encoding the weights in vulnerable layers using a low-collision hash function. During DNN inference, new hashes are extracted from the target layers and compared against the ground-truth signatures. AccHashtag incorporates a lightweight methodology that allows for real-time fault detection on embedded platforms. We devise a specialized compute core for AccHashtag on field-programmable gate arrays (FPGAs) to facilitate online hash generation in parallel to DNN execution. Extensive evaluations with the state-of-the-art bit-flip attack on various DNNs demonstrate the competitive advantage of AccHashtag in terms of both attack detection and execution overhead.

27 January 2022 in IEEE Sensors Journal
Verified
Journal Article
Closed access
SASaehyun Ahn
Saehyun Ahn
JCJung-Woo Chang
Jung-Woo Chang
HYHyeon-Seok Yoon
Hyeon-Seok Yoon
SKSuk-Ju Kang
Suk-Ju Kang
The multi-touch algorithm in touch screens is one of the most important interaction tasks for functional operations. This algorithm classifies multiple objects and recognizes the touch positions. However, rule-based algorithms have limited detection performance for small and closely-spaced touch objects. Deep learning-based multi-touch algorithms achieve higher accuracy than conventional rule-based methods. However, the implementation of state-of-the-art multi-touch algorithms in hardware faces the critical problem that the latency and necessary hardware resources are greatly increased owing to the significant computational complexity. Neural architecture search (NAS) has recently been recognized as an effective solution to search for optimal deep learning models in hardware. This paper proposes an efficient NAS algorithm, TouchNAS, to find the optimal multi-touch algorithm on a target device. We first introduce the progressive architecture pruning method for the search space. This method reduces the training and evaluation costs by omitting the accuracy predictor. We also propose a dynamic learning rate scheduler that assists the efficient training of multiple models by calculating the number of model parameters. Finally, we present multi-touch datasets to measure the performance of multi-touch algorithms objectively. The experimental results on various deep learning models show that our TouchNAS identifies only the candidates for the optimal model within the entire search space for selection and training. Moreover, the optimal model found by TouchNAS achieves superior performance over state-of-the-art NAS methods.