Compact Network Research Articles

Cold argon plasma-modified pea protein isolate: A strategy to enhance ink performance and digestibility in 3D-printed plant-based meat.

Cold argon plasma-modified pea protein isolate: A strategy to enhance ink performance and digestibility in 3D-printed plant-based meat.

Mechanistic study of ultrasound synergy with soybean 11S globulin to improve myofibrillar protein gel properties in low-salt lamb: molecular conformation and water migration.

Mechanistic study of ultrasound synergy with soybean 11S globulin to improve myofibrillar protein gel properties in low-salt lamb: molecular conformation and water migration.

Pixel2Pixel: A Pixelwise Approach for Zero-Shot Single Image Denoising.

We propose Pixel2Pixel, a novel zero-shot image denoising framework that leverages the non-local self-similarity of images to generate a large number of training samples using only the input noisy image. This framework employs a compact convolutional neural network architecture to achieve high-quality image denoising. Given a single observed noisy image, we first aim to obtain multiple images with different noise versions. We ensure that the content remains as consistent as possible with the true signal of the noisy image while keeping the noise independent. Specifically, we construct a pixel bank tensor, where each pixel consists of the most similar pixels from the non-local region of the noisy image. Then, multiple training samples, also known as pseudo instances, can be derived from the pixel bank by randomly pixel sampling. By harnessing pixel-wise random sampling, Pixel2Pixel generates a large number of training pseudo instances, thus avoiding reliance on specific training data. In addition, this non-local pixel selection and random sampling strategy helps to break down the spatial correlation of real-world noise as well. Since the proposed method does not require accurate priors on the noise distribution and clean training images, it is suitable for a wide range of noise types and different noise levels, exhibiting strong generalization ability, especially in real noisy scenes. Extensive experiments across various noise types show that Pixel2Pixel outperforms existing methods.

Enabling Ultralow-Latency Services With Ubiquitous Mobility by Means of a Compact Network Architecture

Enabling Ultralow-Latency Services With Ubiquitous Mobility by Means of a Compact Network Architecture

Reversible engineering of cell membrane receptors based on host-guest recognition for on-demand regulation of cellular behavior.

Reversible engineering of cell membrane receptors based on host-guest recognition for on-demand regulation of cellular behavior.

Lightweight Voice Authentication for IoT Devices Using MFCC and CNN on Edge Hardware

Abstract: This paper presents a lightweight, real-time voice authentication system designed for IoT devices to enhance security by allowing only authorized voice commands. It utilizes a MEMS-based INMP441 microphone and an ESP32-S3 microcontroller to capture audio, extract features via Mel-Frequency Cepstral Coefficients (MFCC), and classify them using a compact Convolutional Neural Network (CNN). Trained with TensorFlow and deployed with TensorFlowLite, the system supports efficient on-device inference, ideal for resource-limited edge hardware.Combining signal processing with deep learning, the solution ensures low latency, minimal power consumption, and enhanced privacy by performing all processing locally—avoiding cloud dependency. It demonstrates robust performance in diverse acoustic conditions and is well-suited for applications in smart homes, healthcare, and industrial automation.This work highlights the viability of embedded AI for secure, intuitive voice interfaces in IoT. Future improvements may include adaptive learning, multi-user support, and integration with other biometric modalities.

VMMCD: VMamba-Based Multi-Scale Feature Guiding Fusion Network for Remote Sensing Change Detection

Remote sensing image change detection, being a pixel-level dense prediction task, requires both high speed and high accuracy. The redundancy within the models and detection errors, particularly missed detections, generally affect accuracy and merit further research. Moreover, the former also leads to a reduction in speed. To guarantee the efficiency of change detection, encompassing both speed and accuracy, a VMamba-based Multi-scale Feature Guiding Fusion Network (VMMCD) is proposed. This network is capable of promptly modeling global relationships and realizing multi-scale feature interaction. Specifically, the Mamba backbone is adopted to replace the commonly used CNN and Transformer backbones. By leveraging VMamba’s global modeling ability with linear computational complexity, the computational resources needed for extracting global features are reduced. Secondly, considering the characteristics of the VMamba model, a compact and efficient lightweight network architecture is devised. The aim is to reduce the model’s redundancy, thereby avoiding the extraction or introduction of interfering and redundant information. As a result, the speed and accuracy of the model are both enhanced. Finally, the Multi-scale Feature Guiding Fusion (MFGF) module is developed, which strengthens the global modeling ability of VMamba. Additionally, it enriches the interaction among multi-scale features to address the common issue of missed detections in changed areas. The proposed network achieves competitive results on three publicly available datasets—SYSU-CD, WHU-CD, and S2Looking—and surpasses the current state-of-the-art (SOTA) methods on the SYSU-CD dataset, with an F1 of 83.35% and IoU of 71.45%. Moreover, for inputs of 256×256 size, it is more than three times faster than the current SOTA VMamba-based change detection model. This outstanding achievement demonstrates the effectiveness of our proposed approach.

Achieving Low-Latency, High-Throughput Online Partial Particle Identification for the NA62 Experiment Using FPGAs and Machine Learning

FPGA-RICH is an FPGA-based online partial particle identification system for the NA62 experiment employing AI techniques. Integrated between the readout of the Ring Imaging Cherenkov detector (RICH) and the low-level trigger processor (L0TP+), FPGA-RICH implements a fast pipeline to process in real-time the RICH raw hit data stream, producing trigger primitives containing elaborate physics information—e.g., the number of charged particles in a physics event—that L0TP+ can use to improve trigger decision efficiency. Deployed on a single FPGA, the system combines classical online processing with a compact Neural Network algorithm to achieve efficient event classification while managing the challenging ∼10 MHz throughput requirement of NA62. The streaming pipeline ensures ∼1 μs latency, comparable to that of the NA62 detectors, allowing its seamless integration in the existing TDAQ setup as an additional detector. Development leverages High-Level Synthesis (HLS) and the open-source hls4ml package software–hardware codesign workflow, enabling fast and flexible reprogramming, debugging, and performance optimization. We describe the implementation of the full processing pipeline, the Neural Network classifier, their functional validation, performance metrics and the system’s current status and outlook.

Multiple‐Asymmetric Molecular Engineering Enables Regio‐Regular Selenium‐Substituted Acceptor with High Efficiency and Ultra‐Low Energy Loss in Binary Organic Solar Cells

Asymmetric molecular engineering utilized for developing efficient small molecular acceptors (SMAs), while adopting multiple asymmetric strategies at the terminals, side chains, and cores of efficient SMAs remains a challenge and effects on reducing energy loss (Eloss) have been rarely investigation. Herein, four regio‐regular multiple‐asymmetric SMAs (DASe‐4F, DASe‐4Cl, TASe‐2Cl2F, and TASe‐2F2Cl) are constructed by delicately manipulating the number and position of F and Cl on end groups. Triple‐asymmetric TASe‐2F2Cl not only exhibits a unique and most compact 3D network crystal stacking structure, but also possesses excellent crystallinity and electron mobility in neat film. Surprisingly, the PM1:TASe‐2F2Cl‐based binary Organic solar cells (OSCs) yield a champion power conversion efficiencies (PCEs) of 19.32%, surpassing the PCE of 18.27%, 17.25% and 16.30% for DASe‐4F, DASe‐4Cl, and TASe‐2Cl2F‐based devices, which attributed to the optimized blend morphology with proper phase separation and more ordered intermolecular stacking, excellent charge transport. Notably, the champion PCE of 19.32% with ultra‐low non‐radiative recombination energy loss (ΔE3) of 0.179 eV marks a record‐breaking result for selenium‐containing SMAs in binary OSCs. Our innovative multiple‐asymmetric molecular engineering of precisely modulating the number and position of fluorinated/chlorinated end groups is an effective strategy for obtaining highly‐efficient and minimal ΔE3 of selenium‐substituted SMAs‐based binary OSCs simultaneously.

Multiple-Asymmetric Molecular Engineering Enables Regioregular Selenium-Substituted Acceptor with High Efficiency and Ultra-low Energy Loss in Binary Organic Solar Cells.

Asymmetric molecular engineering is utilized for developing efficient small molecular acceptors (SMAs), whereas adopting multiple asymmetric strategies at the terminals, side chains, and cores of efficient SMAs remains a challenge, and effects on reducing energy loss (Eloss) have been rarely investigation. Herein, four regioregular multiple-asymmetric SMAs (DASe-4F, DASe-4Cl, TASe-2Cl2F, and TASe-2F2Cl) are constructed by delicately manipulating the number and position of F and Cl on end groups. Triple-asymmetric TASe-2F2Cl not only exhibits a unique and most compact 3D network crystal stacking structure but also possesses excellent crystallinity and electron mobility in neat film. Surprisingly, the PM1:TASe-2F2Cl-based binary organic solar cells (OSCs) yield a champion power conversion efficiencies (PCEs) of 19.32%, surpassing the PCE of 18.27%, 17.25%, and 16.30% for DASe-4F, DASe-4Cl, and TASe-2Cl2F-based devices, which attributed to the optimized blend morphology with proper phase separation and more ordered intermolecular stacking and excellent charge transport. Notably, the champion PCE of 19.32% with ultralow nonradiative recombination energy loss (ΔE3) of 0.179eV marks a record-breaking result for selenium-containing SMAs in binary OSCs. Our innovative multiple-asymmetric molecular engineering of precisely modulating the number and position of fluorinated/chlorinated end groups is an effective strategy for obtaining highly-efficient and minimal ΔE3 of selenium-substituted SMAs-based binary OSCs simultaneously.

Structural and gelation properties of soy protein isolates-Sesbania gum gels: effects of ultrasonic pretreatment and CaSO4 concentration.

Structural and gelation properties of soy protein isolates-Sesbania gum gels: effects of ultrasonic pretreatment and CaSO4 concentration.

DCNN-SBiL: EEG signal based mild cognitive impairment classification using compact convolutional network

DCNN-SBiL: EEG signal based mild cognitive impairment classification using compact convolutional network

Composite cold-set gels of kidney bean protein isolate and basil seed gum induced by glucono-δ-lactone and sodium citrate: Preparation, gel properties and protection on astaxanthin.

Composite cold-set gels of kidney bean protein isolate and basil seed gum induced by glucono-δ-lactone and sodium citrate: Preparation, gel properties and protection on astaxanthin.

Digit-Serial DA-Based Fixed-Point RNNs: A Unified Approach for Enhancing Architectural Efficiency.

The next crucial step in artificial intelligence involves integrating neural network models into embedded and mobile systems. This requires designing compact and energy-efficient neural network models in silicon for optimized performance. This article introduces a unified approach for enhancing the architectural efficiency of long short-term memory (LSTM) recurrent neural networks (RNNs). Precisely, two new structures (I and II) based on the two's complement (TC) digit-serial distributed arithmetic (DSDA) technique are presented. The block-circulant matrix-vector multiplications (MVMs) and element-wise multiplications (EWMs) are formulated using TC DSDA. In addition, a fixed-point (FxP) training procedure for quantized LSTM RNNs is considered and validated for speech recognition tasks. Both structures leverage the circular rotation of weights and generate partial products with input digit slices. A new partial-product generator (PPG) and partial-product selector (PPS) designed to work with both unsigned and signed digits is introduced. In Structure I, a nonpipelined MVM is realized with a few PPGs and PPSs, followed by a shift-accumulate unit (SAU). Conversely, in Structure II, a suitably chosen depth-pipelined MVM is achieved with multiple PPGs and PPSs, followed by a shift-to-add tree (SAT). A critical path delay (CPD) analysis for both the proposed structures is also presented. Compared with previous works, post-synthesis results on 28-nm fully depleted silicon-on-insulator (FDSOI) technology reveal that for a model size of $128 \times 128$ , Structures I and II provide 39.87%, 95.63%, and 30.95%, 91.18% more area and energy efficiencies, respectively.

Characterization of heat-induced whey protein-Dendrobium officinale polysaccharide and its application in goat milk yogurt.

Characterization of heat-induced whey protein-Dendrobium officinale polysaccharide and its application in goat milk yogurt.

Fabrication and characterization of algal oil-loaded Pickering emulsion gels stabilized by whey protein isolate/starch complex as an emergency food.

Fabrication and characterization of algal oil-loaded Pickering emulsion gels stabilized by whey protein isolate/starch complex as an emergency food.

Gallic acid-functionalized chitosan composite for efficient removal of hexavalent chromium in aqueous.

Gallic acid-functionalized chitosan composite for efficient removal of hexavalent chromium in aqueous.

Optimizing Bay Scallop (Argopecten irradians) Product Quality: Moderate Freezing as an Effective Strategy for Improving Adductor Muscle Gel Properties.

The bay scallop (Argopecten irradians) adductor is an attractive raw material for the production of surimi-like products. The gelling properties of raw materials directly affect the quality of surimi-like products. To assess the potential of processing frozen bay scallop adductors into surimi-like products, the effects of short-term freezing treatment on the endogenous transglutaminase (TGase) activity, myofibrillar protein (MP) structure and gelling properties of bay scallop adductors were investigated during 14 days of frozen storage (-18 °C). The results showed that TGase activity in adductor muscles increased significantly during the first 7 days. After 7-14 days, the carbonyl and sulfhydryl contents of the MPs notably changed (increased then decreased). The β-turn content of the MPs increased, indicating stretching and flexibility. Surface hydrophobicity, fluorescence intensity and sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis demonstrated changes in the tertiary structure of the MPs. Compared with gels from fresh samples, gels from scallop adductors frozen for 1 day presented significantly better texture characteristics (breaking force, gel strength, hardness, springiness, cohesiveness, chewiness) and higher water-holding capacity (p < 0.05). However, these properties significantly decreased on the 7th and 14th days (p < 0.05). Microstructural analysis revealed a more compact gel network from 1-day-frozen adductor muscles. These changes in TGase activity and MP structure are key factors influencing the gelling properties of frozen bay scallop adductors. This study provides new insights for improving gel properties during the frozen storage of bay scallop adductors.

Dynamic Contrastive Knowledge Distillation for Efficient Image Restoration

Knowledge distillation (KD) is a valuable yet challenging approach that enhances a compact student network by learning from a high-performance but cumbersome teacher model. However, previous KD methods for image restoration overlook the state of the student during the distillation, adopting a fixed solution space that limits the capability of KD. Additionally, relying solely on L1-type loss struggles to leverage the distribution information of images. In this work, we propose a novel dynamic contrastive knowledge distillation (DCKD) framework for image restoration. Specifically, we introduce dynamic contrastive regularization to perceive the student's learning state and dynamically adjust the distilled solution space using contrastive learning. Additionally, we also propose a distribution mapping module to extract and align the pixel-level category distribution of the teacher and student models. Note that the proposed DCKD is a structure-agnostic distillation framework, which can adapt to different backbones and can be combined with methods that optimize upper-bound constraints to further enhance model performance. Extensive experiments demonstrate that DCKD significantly outperforms the state-of-the-art KD methods across various image restoration tasks and backbones.

Hybrid Data-Free Knowledge Distillation

Data-free knowledge distillation aims to learn a compact student network from a pre-trained large teacher network without using the original training data of the teacher network. Existing collection-based and generation-based methods train student networks by collecting massive real examples and generating synthetic examples, respectively. However, they inevitably become weak in practical scenarios due to the difficulties in gathering or emulating sufficient real-world data. To solve this problem, we propose a novel method called Hybrid Data-Free Distillation (HiDFD), which leverages only a small amount of collected data as well as generates sufficient examples for training student networks. Our HiDFD comprises two primary modules, i.e., the teacher-guided generation and student distillation. The teacher-guided generation module guides a Generative Adversarial Network (GAN) by the teacher network to produce high-quality synthetic examples from very few real-world collected examples. Specifically, we design a feature integration mechanism to prevent the GAN from overfitting and facilitate the reliable representation learning from the teacher network. Meanwhile, we drive a category frequency smoothing technique via the teacher network to balance the generative training of each category. In the student distillation module, we explore a data inflation strategy to properly utilize a blend of real and synthetic data to train the student network via a classifier-sharing-based feature alignment technique. Intensive experiments across multiple benchmarks demonstrate that our HiDFD can achieve state-of-the-art performance using 120 times less collected data than existing methods.