JLJiaming Liu

Jiaming Liu

*This is an algorithmically generated author record based on publicly available information.

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


Jiaming Liu has a total of 68 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • From Tile to Pixel: A Broadcast-Agent Architecture for Continuous Optimization in Streaming Mobile Vision Systems

    by Jiaming Liu, Qi Zheng, Zhijian Hao, Chen Lu, Shuocheng Wang, Ruoxi Zhu, Honglei Chen, Wenzhong Bao, Yibo Fan, Hongkai Xiong

    Edge vision systems in mobile devices, such as drones and autonomous vehicles, rely on sensors and image signal processors to produce high-quality images for human perception and machine vision tasks under varying optical conditions. However, achieving real-time, continuous optimization of vision system remains challenging due to high computational costs and mismatches between sensor data flows and neural network accelerators. This work introduces a pixel-level continuous optimization architecture that integrates a deep reinforcement learning (DRL)-driven broadcast agent with a pixel-granularity streaming computation mode, tailored for resource-constrained mobile platforms. The broadcast agent decomposes bulky DRL models into parallel, lightweight agents with clustered shallow backbones, reducing inference latency. The pixel-level computation mode aligns with sensor raster-scan sequences, minimizing storage overheads through optimized convolution kernels and buffers. Experimental results demonstrate superior performance, achieving 1080P at 75 FPS on FPGA and 240 FPS on ASIC, with reduced hardware resources (1.1106 mm$^{2}$ area, 1.6 W power on 28 nm ASIC) compared to baselines. This approach enables continuous, real-time imaging optimization, advancing the deployment of intelligent edge vision systems in dynamic real-world scenarios.

  • NormMax: A Robust L2-Normalization Activation Function with Precision-Scalable Hardware Architectures for Noisy-Label Learning

    by Chuan Liu, Jiaming Liu, Kexin Wang, Yifan Zhang, Dinghua Wu, Weifan Li, Wenzhong Bao, Honglei Chen, Kexin Liu

    Label noise in real-world datasets poses a significant challenge for deep neural networks, particularly in image classification tasks, where it disrupts learning and degrades performance. The softmax activation function, widely used in Vision Transformers (ViTs), exacerbates this issue through its exponential normalization, leading to gradient instability and noise amplification across layers. Additionally, hardware implementations of softmax introduce quantization errors that further impair high-sensitivity inference under noisy conditions. This paper introduces NormMax, a novel L2-normalization-based activation function designed to enhance robustness against label noise in deep ViTs, paired with two precision-scalable hardware architectures: an iterative, area-optimized design and a fully pipelined, throughput-optimized design with CORDIC-linear and the restoring square root method. Experimental evaluations on CIFAR-10 and CIFAR-100 demonstrate that NormMax achieves a 2.96% accuracy improvement over softmax baselines. The proposed hardware architectures, implemented in GF 28nm ASIC, deliver a mean squared error of 10−14, with the iterative design consuming 0.49 mW and the pipelined design achieving stable ∼2.0 Gops/s in each depth configuration. These advancements enable robust, high-precision inference in noisy-label scenarios, offering significant improvements over traditional softmax-based approaches for both algorithmic performance and hardware efficiency.

  • Open Access

    Accelerating spiking neural networks with photonic reconfigurable devices

    by Chen Lu, Kangli Xu, Jiaming Liu, Yuanhong Ding, Jiajie Yu, Jieru Song, Jialin Meng, Tianyu Wang, Shuming Guo, Xingcheng Jin, et al.

    Spiking neural networks face hardware limitations as conventional architectures exhibit low array utilization, underperforming GPU-driven artificial neural networks in vision tasks. We present a programmable spiking neurocomputing architecture using CMOS-compatible photonic reconfigurable devices that unify synaptic/neuronal functions in single components. The optical isolation of photonic reconfigurable devices suppresses inter-cell crosstalk while enabling independent neuronal/synaptic programmability without operational redundancy. Programmable spiking neurocomputing architecture maximizes array utilization to enhance computational efficiency for spiking neural networks acceleration. Compared with conventional architectures, our architecture on spiking Visual Geometry Group networks demonstrates 1176× latency reduction and 239× energy savings in static recognition/classification and dynamic detection/tracking tasks, while maintaining equivalent recognition accuracy. This photonic-electronic integration establishes a scalable hardware framework that bridges the performance gap between spiking neural networks and state-of-the-art artificial neural networks, particularly for complex visual information processing applications requiring both efficiency and precision.


  • One-Iteration ISP Controller for Real-Time Machine Vision

    by Jiaming Liu, Zhijian Hao, Qi Zheng, Ruoxi Zhu, Shuocheng Wang, Honglei Chen, Wenzhong Bao, Hongkai Xiong, Yibo Fan

    Conventional image signal processing (ISP) control algorithms based on human visual perception are insufficient for the demands of modern machine vision systems. Although recent learning-based methods have attempted to adapt ISP hyperparameters for specific vision tasks, their high latency and hardware cost of iterative optimization hinder deployment on edge devices such as those used in autonomous driving. To address these limitations, this paper proposes a real-time machine vision system through algorithm–hardware co-design. First, we introduce a one-iteration learning framework to avoid iterative optimization, significantly reducing latency for real-time use. Second, we propose a hardware-friendly controller, RasterNet, specifically tailored for raster-scanning sensor dataflow, eliminating redundant computation. Third, we present a pipelined ISP controller architecture incorporating branch and chroma time division multiplexing techniques to minimize the number of processing elements, achieving a compact and efficient design. Experiments demonstrate that the proposed system achieves superior object detection accuracy on resource-constrained platforms, with real-time performance reaching 70 FPS on FPGA and 224 FPS on ASIC implementations.


  • Closely-Coupled Reconstruction and Generation for Blind Face Restoration

    by Shichen Peng, Jiaming Liu, Zihao Liu, Hongzhong Zheng, Dimin Niu, Xiaoyang Zeng, Yibo Fan

    Blind Face Restoration (BFR) involves restoring high-quality images from various unknown and severely degraded counterparts, which is a challenging task due to the conflicting objectives of content reconstruction and detail generation. In this paper, we propose a closely-coupled approach to address this problem and achieve photorealistic and faithful reproductions. Specifically, we propose a two-step image restoration model that consists of the following steps: Firstly, we train a BaseNet that incorporates a filtered feature fusion module ($F^{3}M$) to purify degraded feature maps. Secondly, while keeping the BaseNet fixed, we train a DetailNet that utilizes a feature probabilistic model to generate high-frequency detail information. The proposed framework not only separates the reconstruction and generation processes but also deeply analyzes their interactions, leading to an optimized balance between perceptual quality and fidelity. Our approach is validated through extensive experiments on both synthetic datasets and real-world facial photographs, demonstrating significant improvements in Frechet Inception Distance (FID) scores while maintaining identity consistency. The experimental results highlight our method's state-of-the-art performance, achieving superior visual quality and processing efficiency compared to existing methods.


  • A Dual-Generalization Low-Light Enhancement Framework for Capsule Endoscopy Image Restoration and Segmentation

    by Shuocheng Wang, Jiaming Liu, Ruoxi Zhu, Chengkang Huang, Minge Jing, Yibo Fan

    In recent years, deep learning technology has automated the diagnosis of gastrointestinal (GI) tract disease, enabling doctor-machine collaborative diagnosis. However, the images captured by wireless capsule endoscopy (WCE) easily suffer from varying brightness levels of low-light degradation due to the complex structure of GI tract and the limitations of the light source, which impacts both human and machine diagnostic accuracy. Moreover, images may contain varying degrees of structural and semantic details even under a similar brightness level, which still can compromise segmentation accuracy. To address these issues, we propose a dual-generalization framework for low-light WCE images. Our framework includes an Image Guidance and Laplacian Fusion Module (IGLFM), a Brightness Level Generalization Module (BLGM) and a Wavelet Segmentation Generalization Module (WSGM). IGLFM and BLGM can restore low-light images across different brightness levels and WSGM can enhance segmentation accuracy by generalizing the varying degrees of details across images. With BLGM and WSGM, our framework enables two aspects of generalization: generalization to input images with different brightness levels and generalization to images with varying detail levels. Extensive experiments demonstrate that our method achieves significant performance under varying brightness levels and improvements in segmentation accuracy, surpassing the existing state-of-the-art (SOTA) method with gains of 4.70 dB / 0.022 (PSNR/SSIM) on Kvasir-Capsule dataset and 1.61 dB / 0.018 on RLE dataset. WSGM consistently improves segmentation accuracy across six popular networks, achieving up to +4.7% mIoU and +5.3% Dice improvements on RLE dataset. Our code will be available at https://github.com/superwsc/Dual-Gen-Frame.


  • Open Access

    Unicorn: Unified Neural Image Compression with One Number Reconstruction

    by Qi Zheng, Haozhi Wang, Zihao Liu, Jiaming Liu, Zhijian Hao, Bu Chen, Min Li, Rui Wan, Peiye Liu, Yanheng Lu, et al.


  • A Flexible Zero-Shot Approach to Tone Mapping via Structure-Preserving Diffusion Models

    by Ruoxi Zhu, Shusong Xu, Peiye Liu, Jiaming Liu, Yanheng Lu, Dimin Niu, Hongzhong Zheng, Yen-Kuang Chen, Ming'E Jing, Yibo Fan

    With the prevalence of high dynamic range (HDR) imaging, tone mapping techniques, which convert HDR images to high-quality standard dynamic range (SDR) images for display, have become increasingly important. However, obtaining paired HDR and high-quality SDR images is almost impossible, posing challenges to learning-based tone mapping methods. To address this issue, we propose a zero-shot tone mapping framework without requiring any HDR training samples. Our approach decomposes images into two components: structural information and tonal information. A diffusion-based mapping model taking the structural information as input is first trained in the high-quality SDR domain, then transferred to the HDR domain that has less readily available training data for inference, leveraging the equivalent distribution of the structural information across both domains. To preserve the original image’s structure, we modify the reverse sampling process and explicitly incorporate the original structural information into the intermediate results. To improve the image details, we introduce a dual-control network, enabling different conditional inputs to control different scales of the output. Additionally, we devise a flexible tone adjustment strategy, with a bunch of novel loss functions to modify the trained score function dynamically during reverse sampling, allowing users to customize the style of the generated image according to their preference during testing. Initially designed for tone mapping, our model can be applied to various tasks including image fusion, exposure correction, dehazing, etc., without retraining. Experimental results demonstrate that our approach surpasses previous state-of-the-art methods, indicating that it can serve as an effective, flexible and versatile solution to various tone-mapping tasks. Source code is available at https://github.com/ZSDM-HDR/Zero-Shot-Diffusion-HDR.