Junha Ryu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Junha Ryu has a total of 39 co-authors. Here are the top co-authors by number of shared publications:
- Hoi-Jun Yoo
School of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, South Korea
16 shared publications - Donghyeon Han
Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology (MIT), Cambridge, MA, USA
10 shared publications - Gwangtae Park
School of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, South Korea
9 shared publications - Dongseok Im
School of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, South Korea
8 shared publications - Sanghoon Kang
School of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, South Korea
5 shared publications - Sangyeob Kim
School of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, South Korea
5 shared publications
Latest Publications
- 22 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 34 (1) , 118-129
An Energy-Efficient High Resolution Vision Transformer Processor Exploiting Token Similarity Beyond Token Merging
Token merging is a widely adopted technique in vision transformers (ViTs) to reduce computation with minimal accuracy degradation. However, in tasks that require fine-grained outputs, such as text-unified object recognition, aggressive merging reaches a limit due to accuracy loss. To address this, we introduce a method that exploits residual token similarity beyond the merging threshold using difference computing, which significantly improves energy efficiency. The proposed processor integrates the post merging residual difference (PMRD) algorithm, which first reduces 34% of tokens through merging, then applies difference computing to the remaining tokens. This results in a 31% reduction in effective bit operations by leveraging residual similarity. To further exploit the resulting bit-level sparsity, the processor employs sparsity-separated accumulation with sign-magnitude representation (SSA-SM) and a bi-directional dynamic memory allocation (BDA) to optimize both computation and memory access. These enhancements yield an additional 29.4% computational energy savings. Overall, the system achieves a $2.22\times $ improvement in energy efficiency and delivers 7.23 TOPS/W on the MS COCO dataset for text-unified object recognition.
- 22 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 34 (1) , 118-129
An Energy-Efficient High Resolution Vision Transformer Processor Exploiting Token Similarity Beyond Token Merging
Token merging is a widely adopted technique in vision transformers (ViTs) to reduce computation with minimal accuracy degradation. However, in tasks that require fine-grained outputs, such as text-unified object recognition, aggressive merging reaches a limit due to accuracy loss. To address this, we introduce a method that exploits residual token similarity beyond the merging threshold using difference computing, which significantly improves energy efficiency. The proposed processor integrates the post merging residual difference (PMRD) algorithm, which first reduces 34% of tokens through merging, then applies difference computing to the remaining tokens. This results in a 31% reduction in effective bit operations by leveraging residual similarity. To further exploit the resulting bit-level sparsity, the processor employs sparsity-separated accumulation with sign-magnitude representation (SSA-SM) and a bi-directional dynamic memory allocation (BDA) to optimize both computation and memory access. These enhancements yield an additional 29.4% computational energy savings. Overall, the system achieves a $2.22\times $ improvement in energy efficiency and delivers 7.23 TOPS/W on the MS COCO dataset for text-unified object recognition.
- 2 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 72 (11) , 1610-1614
A 227.1 TOPS/W High Energy-Efficiency PNN-Based 3-D Object Recognition Processor With Spiking Neural Network for Edge Device
This paper proposes a high-efficiency spiking neural network (SNN) processor for point-cloud neural network (PNN) based 3D object recognition on edge devices. PNN independently process points, leading to redundant computations and low energy efficiency. SNN leverage the characteristics of accumulation operations and the properties of Leaky Integrate-and-Fire (LIF) activation, to optimize efficiency. A novel 3D object recognition processor maximizes the characteristics of SNN by introducing two key features: 1) intra-timestep computation skipping and 2) timestep reduction via inter-timestep correlation analysis. The Event-driven Bit-serial Processing Unit (EBPU) utilizes intratimestep computation skipping for up to 7.2× higher energyefficiency. The Output Spike Precognition Unit (OSPU) enhances inter-timestep correlation analysis for 65% improvement in throughput. Consequently, the proposed processor achieves an energy efficiency of 227.1 TOPS/W for 3D object classification, 218.1 TOPS/W for part segmentation, and 92.5 TOPS/W for semantic segmentation, achieving up to 22.9× higher energy efficiency than the conventional state-of-the-art PNN accelerators
- 2 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 72 (11) , 1610-1614
A 227.1 TOPS/W High Energy-Efficiency PNN-Based 3-D Object Recognition Processor With Spiking Neural Network for Edge Device
This paper proposes a high-efficiency spiking neural network (SNN) processor for point-cloud neural network (PNN) based 3D object recognition on edge devices. PNN independently process points, leading to redundant computations and low energy efficiency. SNN leverage the characteristics of accumulation operations and the properties of Leaky Integrate-and-Fire (LIF) activation, to optimize efficiency. A novel 3D object recognition processor maximizes the characteristics of SNN by introducing two key features: 1) intra-timestep computation skipping and 2) timestep reduction via inter-timestep correlation analysis. The Event-driven Bit-serial Processing Unit (EBPU) utilizes intratimestep computation skipping for up to 7.2× higher energyefficiency. The Output Spike Precognition Unit (OSPU) enhances inter-timestep correlation analysis for 65% improvement in throughput. Consequently, the proposed processor achieves an energy efficiency of 227.1 TOPS/W for 3D object classification, 218.1 TOPS/W for part segmentation, and 92.5 TOPS/W for semantic segmentation, achieving up to 22.9× higher energy efficiency than the conventional state-of-the-art PNN accelerators
- Open Access
- 5 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Solid-State Circuits Society
- Vol. 5, 212-228
An Overview of AI Hardware Architectures and Silicon for 3-D Spatial Computing Systems
As artificial intelligence (AI) advances, 3D spatial computing has emerged as a key application in various fields. It interprets the 3D space surrounding users and provides them with useful information. This paper presents a survey of AI hardware architectures and silicon solutions for 3D spatial computing systems. The survey categorizes five domains: 3D data capturing, 3D data analysis, 3D hand motion analysis, simultaneous localization and mapping (SLAM), and 3D rendering. Each session analyzes design considerations for domain-specific accelerators. Finally, the paper discusses a next-generation 3D spatial computing platform that integrates various functions of 3D spatial computing systems using AI technologies.
- 5 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Solid-State Circuits Society
- Vol. 5, 212-228
Open AccessAn Overview of AI Hardware Architectures and Silicon for 3-D Spatial Computing Systems
As artificial intelligence (AI) advances, 3D spatial computing has emerged as a key application in various fields. It interprets the 3D space surrounding users and provides them with useful information. This paper presents a survey of AI hardware architectures and silicon solutions for 3D spatial computing systems. The survey categorizes five domains: 3D data capturing, 3D data analysis, 3D hand motion analysis, simultaneous localization and mapping (SLAM), and 3D rendering. Each session analyzes design considerations for domain-specific accelerators. Finally, the paper discusses a next-generation 3D spatial computing platform that integrates various functions of 3D spatial computing systems using AI technologies.
- 30 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 60 (11) , 4154-4166
NeRF-Navi: An Energy-Efficient NeRF 3-D Path Planning Processor With Reconfigurable Approximate/Accurate Bit Offloading Core
An implicit neural representation (INR) continuously encodes a 3-D space using a neural network. Neural radiance field (NeRF), a type of INR, achieves a high path planning success rate of 98.6%. It leverages the continuous space representation ability of NeRF. However, accelerating NeRF path planning on edge devices faces limitations due to the excessive computational load. In this article, we present NeRF-Navi, an accurate and energy-efficient 3-D NeRF path planning processor with three key features: 1) dual-attention neural path sampling (DANPS) engine uses map and collision attention to reduce the number of redundant batches, which saves 96.2% of system energy; 2) approximate-accurate (A2) core with an error-compensable reduction tree (ECRT) introduces three approximate computing modes with less than 1.6% accuracy overhead; bit sparsity boosting logic (BSBL) increases bit sparsity and reduces the compensation overhead of ECRT; and the A2 core and BSBL achieve a 26.2% reduction in total system energy; and 3) outlier channel bit-offloading core and bit allocator (BA) that offload sparse MSB bits of outlier channels, reducing total system energy by 36%. NeRF-Navi is fabricated in a 28-nm logic CMOS process and occupies a 6.48-mm2 die area. NeRF-Navi finally achieves 1.2–8.8× lower path planning energy per task and 1.7–124× lower EDP than the previous 3-D path planning processor.
- 30 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 60 (11) , 4154-4166
NeRF-Navi: An Energy-Efficient NeRF 3-D Path Planning Processor With Reconfigurable Approximate/Accurate Bit Offloading Core
An implicit neural representation (INR) continuously encodes a 3-D space using a neural network. Neural radiance field (NeRF), a type of INR, achieves a high path planning success rate of 98.6%. It leverages the continuous space representation ability of NeRF. However, accelerating NeRF path planning on edge devices faces limitations due to the excessive computational load. In this article, we present NeRF-Navi, an accurate and energy-efficient 3-D NeRF path planning processor with three key features: 1) dual-attention neural path sampling (DANPS) engine uses map and collision attention to reduce the number of redundant batches, which saves 96.2% of system energy; 2) approximate-accurate (A2) core with an error-compensable reduction tree (ECRT) introduces three approximate computing modes with less than 1.6% accuracy overhead; bit sparsity boosting logic (BSBL) increases bit sparsity and reduces the compensation overhead of ECRT; and the A2 core and BSBL achieve a 26.2% reduction in total system energy; and 3) outlier channel bit-offloading core and bit allocator (BA) that offload sparse MSB bits of outlier channels, reducing total system energy by 36%. NeRF-Navi is fabricated in a 28-nm logic CMOS process and occupies a 6.48-mm2 die area. NeRF-Navi finally achieves 1.2–8.8× lower path planning energy per task and 1.7–124× lower EDP than the previous 3-D path planning processor.
- Open Access
- 17 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
- Vol. 15 (2) , 299-311
An Overview of Neural Rendering Accelerators: Challenges, Trends, and Future Directions
Rapid advancements in neural rendering have revolutionized the fields of augmented reality (AR) and virtual reality (VR) by enabling photorealistic 3D modeling and rendering. However, deploying neural rendering on edge devices presents significant challenges due to computational complexity, memory inefficiencies, and energy constraints. This paper provides a comprehensive overview of neural rendering accelerators, identifying the major hardware inefficiencies across sampling, positional encoding, and multi-layer perception (MLP) stages. We explore hardware-software co-optimization techniques that address these challenges and provide a summary for in-depth analysis. Additionally, emerging trends like 3D Gaussian Splatting (3DGS) and hybrid rendering approaches are briefly introduced, highlighting their potential to improve rendering quality and efficiency. By presenting a unified analysis of challenges, solutions, and future directions, this work aims to guide the development of next-generation neural rendering accelerators, especially for resource-constrained environments.
- 17 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
- Vol. 15 (2) , 299-311
Open AccessAn Overview of Neural Rendering Accelerators: Challenges, Trends, and Future Directions
Rapid advancements in neural rendering have revolutionized the fields of augmented reality (AR) and virtual reality (VR) by enabling photorealistic 3D modeling and rendering. However, deploying neural rendering on edge devices presents significant challenges due to computational complexity, memory inefficiencies, and energy constraints. This paper provides a comprehensive overview of neural rendering accelerators, identifying the major hardware inefficiencies across sampling, positional encoding, and multi-layer perception (MLP) stages. We explore hardware-software co-optimization techniques that address these challenges and provide a summary for in-depth analysis. Additionally, emerging trends like 3D Gaussian Splatting (3DGS) and hybrid rendering approaches are briefly introduced, highlighting their potential to improve rendering quality and efficiency. By presenting a unified analysis of challenges, solutions, and future directions, this work aims to guide the development of next-generation neural rendering accelerators, especially for resource-constrained environments.
- 16 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 60 (1) , 99-111
NeuGPU: An Energy-Efficient Neural Graphics Processing Unit for Instant Modeling and Real-Time Rendering on Mobile Devices
Neural radiance field (NeRF) is an emerging computer graphics task that is used for 3-D modeling and rendering in the metaverse, providing a user-friendly and immersive experience. It successfully replaces conventional 3-D modeling methods such as photogrammetry thanks to its simplicity, photo-realistic rendering quality, and versatility. However, it has limitations to be accelerated on mobile augmented reality (AR)/virtual reality (VR) devices due to its memory-intensive hash encoding and extensive computational load. Previous NeRF application specified integrated circuits (ASICs) only supported high frame-rate rendering while being incapable of immediate 3-D model creation. In this article, we present neural graphics processing unit (NeuGPU) to achieve both instant 3-D modeling and real-time rendering using NeRF with three key features. First, segmented hashing with spatial pruning (SHSP) partitions hash table (HT) into multiple segments and reduces 66% of external memory access (EMA). Spatial management unit (SMU) supports SHSP operation through three types of sub-block management. Second, attention-based hybrid interpolation unit (AHIU) optimizes heterogeneous memory access characteristics of hash-based NeRF, achieving 56.4% power reduction. Third, similarity–sparsity skipping (S 3 ) core supports energy-efficient multi-layer perceptron (MLP) operation by applying coarse-grained skip for both similar and sparse data. As a result, NeuGPU is fabricated in a 28-nm process and occupies a 20.25-mm 2 die area. Evaluated on the synthetic NeRF dataset, NeuGPU finally achieves 8.7 $\times$ faster modeling and 231.4 $\times$ smaller energy per iteration compared with edge GPU even though it has 31 $\times$ smaller external memory bandwidth.
- 16 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 60 (1) , 99-111
NeuGPU: An Energy-Efficient Neural Graphics Processing Unit for Instant Modeling and Real-Time Rendering on Mobile Devices
Neural radiance field (NeRF) is an emerging computer graphics task that is used for 3-D modeling and rendering in the metaverse, providing a user-friendly and immersive experience. It successfully replaces conventional 3-D modeling methods such as photogrammetry thanks to its simplicity, photo-realistic rendering quality, and versatility. However, it has limitations to be accelerated on mobile augmented reality (AR)/virtual reality (VR) devices due to its memory-intensive hash encoding and extensive computational load. Previous NeRF application specified integrated circuits (ASICs) only supported high frame-rate rendering while being incapable of immediate 3-D model creation. In this article, we present neural graphics processing unit (NeuGPU) to achieve both instant 3-D modeling and real-time rendering using NeRF with three key features. First, segmented hashing with spatial pruning (SHSP) partitions hash table (HT) into multiple segments and reduces 66% of external memory access (EMA). Spatial management unit (SMU) supports SHSP operation through three types of sub-block management. Second, attention-based hybrid interpolation unit (AHIU) optimizes heterogeneous memory access characteristics of hash-based NeRF, achieving 56.4% power reduction. Third, similarity–sparsity skipping (S 3 ) core supports energy-efficient multi-layer perceptron (MLP) operation by applying coarse-grained skip for both similar and sparse data. As a result, NeuGPU is fabricated in a 28-nm process and occupies a 20.25-mm 2 die area. Evaluated on the synthetic NeRF dataset, NeuGPU finally achieves 8.7 $\times$ faster modeling and 231.4 $\times$ smaller energy per iteration compared with edge GPU even though it has 31 $\times$ smaller external memory bandwidth.
- 11 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 71 (5) , 2529-2533
A 8.81 TFLOPS/W Deep-Reinforcement-Learning Accelerator With Delta-Based Weight Sharing and Block-Mantissa Reconfigurable PE Array
TD3 is one of the most high-performing Deep Reinforcement Learning (DRL) algorithms, providing high training stability and rewards. However, it suffers from low energy efficiency due to high External Memory Access (EMA) and floating point operations. To mitigate this issue and achieve higher throughput and energy efficiency, we propose the DRL accelerator with 3 features: 1) Delta-based Weight Sharing (DWS) represents weights by referencing corresponding network and exploits data locality, leading to EMA reduction of up to 64.3% in feed-forward stage and 39.7% in gradient generation and weight update stage. 2) Block-Mantissa Reconfigurable PE Array (BMRPA) supports variable operations in blocks and mantissa to provide optimal precision for each layer, resulting in up to a 4× increase in throughput. 3) Multi-mode Data Fetcher (MDF) supports bit width adaptive data fetching, achieving twice the bandwidth with an average read overhead of 5.3%. When combined with BMRPA, it attains an energy efficiency of 8.81 TFLOPS/W.
- 11 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 71 (5) , 2529-2533
A 8.81 TFLOPS/W Deep-Reinforcement-Learning Accelerator With Delta-Based Weight Sharing and Block-Mantissa Reconfigurable PE Array
TD3 is one of the most high-performing Deep Reinforcement Learning (DRL) algorithms, providing high training stability and rewards. However, it suffers from low energy efficiency due to high External Memory Access (EMA) and floating point operations. To mitigate this issue and achieve higher throughput and energy efficiency, we propose the DRL accelerator with 3 features: 1) Delta-based Weight Sharing (DWS) represents weights by referencing corresponding network and exploits data locality, leading to EMA reduction of up to 64.3% in feed-forward stage and 39.7% in gradient generation and weight update stage. 2) Block-Mantissa Reconfigurable PE Array (BMRPA) supports variable operations in blocks and mantissa to provide optimal precision for each layer, resulting in up to a 4× increase in throughput. 3) Multi-mode Data Fetcher (MDF) supports bit width adaptive data fetching, achieving twice the bandwidth with an average read overhead of 5.3%. When combined with BMRPA, it attains an energy efficiency of 8.81 TFLOPS/W.
- 3 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 44 (1) , 17-27
A Low-Power Artificial-Intelligence-Based 3-D Rendering Processor With Hybrid Deep Neural Network Computing
A low-power artificial intelligence (AI)-based 3D rendering processor is proposed for metaverse solutions in mobile platforms. It suggests a brain-inspired rendering acceleration architecture designed with a visual perception core. It removes useless computations by realizing 1) spatial attention, 2) temporal familiarity, and 3) top-down attention. The remaining deep neural network (DNN) inference tasks are accelerated by a hybrid neural engine that utilizes both coarse-grained and fine-grained sparsity exploitation simultaneously. It divides the DNN tasks as sparse and dense data and allocates them to the two different neural engines which focus on zero-skipping and data reusability respectively. Thanks to the centrifugal sampling-based workload prediction, it can dynamically divide DNN computations while minimizing PSNR loss caused by the prediction. Fabricated with 28-nm CMOS technology, the processor successfully demonstrates a maximum of 118-FPS rendering while consuming 99.95% lower power compared with modern GPUs.
- 3 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 44 (1) , 17-27
A Low-Power Artificial-Intelligence-Based 3-D Rendering Processor With Hybrid Deep Neural Network Computing
A low-power artificial intelligence (AI)-based 3D rendering processor is proposed for metaverse solutions in mobile platforms. It suggests a brain-inspired rendering acceleration architecture designed with a visual perception core. It removes useless computations by realizing 1) spatial attention, 2) temporal familiarity, and 3) top-down attention. The remaining deep neural network (DNN) inference tasks are accelerated by a hybrid neural engine that utilizes both coarse-grained and fine-grained sparsity exploitation simultaneously. It divides the DNN tasks as sparse and dense data and allocates them to the two different neural engines which focus on zero-skipping and data reusability respectively. Thanks to the centrifugal sampling-based workload prediction, it can dynamically divide DNN computations while minimizing PSNR loss caused by the prediction. Fabricated with 28-nm CMOS technology, the processor successfully demonstrates a maximum of 118-FPS rendering while consuming 99.95% lower power compared with modern GPUs.