Sang-Seol Lee
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Sang-Seol Lee has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Sung-Joon Jang
Korea Electronics Technology Institute (KETI), Seongnam-si, South Korea
7 shared publications - Eunchong Lee
Intelligent Image Processing Research Center, Korea Electronics Technology Institute, Seongnam-si 13488, Republic of Korea
5 shared publications - Jonghee Park
Intelligent Image Processing Research Center, Korea Electronics Technology Institute
2 shared publications - Kyungho Kim
Korea Electronics Technology Institute (KETI)
2 shared publications - Minkyu Lee
Korea Electronics Technology Institute (KETI)
2 shared publications - Youngbae Hwang
Department of Intelligent Systems and Robotics, Chungbuk National University, Cheongju, Republic of Korea
2 shared publications
Latest Publications
- 1 September 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 34 (9) , 2818-2831
TetraPIXA: A High-Performance and Resource-Efficient Accelerator for Pillar and Voxel Feature Networks in 3-D Point Cloud Object Detection
- 1 September 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 34 (9) , 2818-2831
TetraPIXA: A High-Performance and Resource-Efficient Accelerator for Pillar and Voxel Feature Networks in 3-D Point Cloud Object Detection
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 1
AWa: An Activation-Weight Distribution-aware Binarization Method for Large Language Models
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 1
Open AccessAWa: An Activation-Weight Distribution-aware Binarization Method for Large Language Models
- Open Access
- 23 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-6
HAIL-DIMM: Host Access Interleaved with Near-Data Processing on DIMM-based Memory System
Near-data processing (NDP), a solution to reduce data movement overhead between host and memory, should not interfere with host access to ensure system fairness. We propose a cost-effective and energy-efficient LRDIMM-based NDP architecture (HAIL-DIMM) that can seamlessly interleave NDP and host access and is a drop-in replacement for existing main memory modules. The proposed NDP exploits the interleaving capability of the memory controller to interleave NDP and host access naturally. To take advantage of bank interleaving, an atomic operation of the proposed NDP, which consists of data movement and computation, is recognized by the memory controller as a DDR READ/WRITE but by the HAIL-DIMM as NDP based on the request's address. We implement a prototype of the proposed NDP architecture on an FPGA platform as proof of concept. The evaluation results show that the NDP system achieves up to 2.19x speedup in latency and up to 45.4% energy saving for data movement over the baseline system for memory-bound workloads.
- 23 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-6
Open AccessHAIL-DIMM: Host Access Interleaved with Near-Data Processing on DIMM-based Memory System
Near-data processing (NDP), a solution to reduce data movement overhead between host and memory, should not interfere with host access to ensure system fairness. We propose a cost-effective and energy-efficient LRDIMM-based NDP architecture (HAIL-DIMM) that can seamlessly interleave NDP and host access and is a drop-in replacement for existing main memory modules. The proposed NDP exploits the interleaving capability of the memory controller to interleave NDP and host access naturally. To take advantage of bank interleaving, an atomic operation of the proposed NDP, which consists of data movement and computation, is recognized by the memory controller as a DDR READ/WRITE but by the HAIL-DIMM as NDP based on the request's address. We implement a prototype of the proposed NDP architecture on an FPGA platform as proof of concept. The evaluation results show that the NDP system achieves up to 2.19x speedup in latency and up to 45.4% energy saving for data movement over the baseline system for memory-bound workloads.
- 26 October 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-3
Quantization Aware Training with Order Strategy for CNN
In this paper, we present a novel quantization-aware training method proceeding in a specific order in deep learning network. Conventional quantization techniques generally perform all layers at the same time. However, quantization sensitivity is different for each layer. For example, a quantization error increases as the quantizing bit-widths decrease, therefore the quantized model is hard to converge in the training process. To handle the issue, we propose the layer selection module which controls the order of layers to be quantized. Specifically, each layer is sequentially quantized in the order of selection so that the quantized model can be converged effectively. In our experiments, we show the quantization results on the CIFAR-10 image dataset. We also analyze the results between the different selection methods and verify the effectiveness of our proposed method.
- 26 October 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-3
Quantization Aware Training with Order Strategy for CNN
In this paper, we present a novel quantization-aware training method proceeding in a specific order in deep learning network. Conventional quantization techniques generally perform all layers at the same time. However, quantization sensitivity is different for each layer. For example, a quantization error increases as the quantizing bit-widths decrease, therefore the quantized model is hard to converge in the training process. To handle the issue, we propose the layer selection module which controls the order of layers to be quantized. Specifically, each layer is sequentially quantized in the order of selection so that the quantized model can be converged effectively. In our experiments, we show the quantization results on the CIFAR-10 image dataset. We also analyze the results between the different selection methods and verify the effectiveness of our proposed method.
- 21 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 302-303
Implementation of a Round Robin Processing Element for Deep Learning Accelerator
The deep learning acceleration hardware performance is greatly affected by Processing Elements (PEs). In order to apply deep learning accelerators to mobile devices, optimized PE must be designed as ASIC. To improve the performance of PE, we focused on methods of minimizing external memory access and parallelization. As a result, a deep learning accelerator architecture consisting of 512 PEs in parallel is proposed and the results of FPGA implementation is presented.
- 21 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 302-303
Implementation of a Round Robin Processing Element for Deep Learning Accelerator
The deep learning acceleration hardware performance is greatly affected by Processing Elements (PEs). In order to apply deep learning accelerators to mobile devices, optimized PE must be designed as ASIC. To improve the performance of PE, we focused on methods of minimizing external memory access and parallelization. As a result, a deep learning accelerator architecture consisting of 512 PEs in parallel is proposed and the results of FPGA implementation is presented.
- 21 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 324-325
Efficient final output feature map processing method supporting real-time object detection and recognition
Recent advanced in CNN acceleration systems using hardware and software co-design for real-time object detection and recognition, the speedup of post processing can reduce the burden of designing the acceleration logic to meet the real-time goal. In this paper, we propose an optimization method that eliminates the operation of acquiring redundant bounding boxes without degrading performance. It is implemented on a Xilinx Zynq Ultrascale+ MPSoC ZCU106 and shows the performance improvement about 20% that of the conventional method.
- 21 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 324-325
Efficient final output feature map processing method supporting real-time object detection and recognition
Recent advanced in CNN acceleration systems using hardware and software co-design for real-time object detection and recognition, the speedup of post processing can reduce the burden of designing the acceleration logic to meet the real-time goal. In this paper, we propose an optimization method that eliminates the operation of acquiring redundant bounding boxes without degrading performance. It is implemented on a Xilinx Zynq Ultrascale+ MPSoC ZCU106 and shows the performance improvement about 20% that of the conventional method.
- 1 October 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 309-310
Hardware implementation of fast high dynamic range processor for real-time 4K UHD video
The high dynamic range (HDR) has become very important because of the rapid increase in demand for a variety of applications. However, most of them were implemented by expensive systems due to the high complex computation for processing the real-time 4K UHD video. In the proposed hardware, the non-linear camera response function (CRF) with the area optimization of logarithmic computations has been applied to improve HDR quality. And, for embedding in Field Programmable Gate Array (FPGA), we implement a dedicated hardware using 4006 lookup table (LUT) and 21KB sized internal memory. The proposed architecture enables a real-time HDR processing with pipelining for a UHD video (8 Mega pixels) at 30 frames per second.
- 1 October 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 309-310
Hardware implementation of fast high dynamic range processor for real-time 4K UHD video
The high dynamic range (HDR) has become very important because of the rapid increase in demand for a variety of applications. However, most of them were implemented by expensive systems due to the high complex computation for processing the real-time 4K UHD video. In the proposed hardware, the non-linear camera response function (CRF) with the area optimization of logarithmic computations has been applied to improve HDR quality. And, for embedding in Field Programmable Gate Array (FPGA), we implement a dedicated hardware using 4006 lookup table (LUT) and 21KB sized internal memory. The proposed architecture enables a real-time HDR processing with pipelining for a UHD video (8 Mega pixels) at 30 frames per second.
- 1 October 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-2
Low-complexity hardware architecture of traffic sign recognition with IHSL color space for advanced driver assistance systems
This paper presents novel hardware architecture with low-complexity color conversion scheme and parallel processing of red region detection for the applications of automatic traffic sign detection system. By the inherent parallelism of the various red region detections, we designed a fully pipelined architecture implemented on the FPGA platform. The proposed architecture enables a real-time traffic sign recognition (TSR) processing for a full-HD video (2 Mega pixels) at 60 frames per second.
- 1 October 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-2
Low-complexity hardware architecture of traffic sign recognition with IHSL color space for advanced driver assistance systems
This paper presents novel hardware architecture with low-complexity color conversion scheme and parallel processing of red region detection for the applications of automatic traffic sign detection system. By the inherent parallelism of the various red region detections, we designed a fully pipelined architecture implemented on the FPGA platform. The proposed architecture enables a real-time traffic sign recognition (TSR) processing for a full-HD video (2 Mega pixels) at 60 frames per second.
- 1 September 1991
- Published by Elsevier in Mechanical Systems and Signal Processing
- Vol. 5 (5) , 403-420
Position of non-linear elements and identification of their type by a local non-parametric method
- 1 September 1991
- Published by Elsevier in Mechanical Systems and Signal Processing
- Vol. 5 (5) , 403-420
Position of non-linear elements and identification of their type by a local non-parametric method