Lightweight 3D Research Articles

A Comprehensive Framework for Transportation Infrastructure Digitalization: TJYRoad-Net for Enhanced Point Cloud Segmentation

This research introduces a cutting-edge approach to traffic infrastructure digitization, integrating UAV oblique photography with LiDAR point clouds for high-precision, lightweight 3D road modeling. The proposed method addresses the challenge of accurately capturing the current state of infrastructure while minimizing redundancy and optimizing computational efficiency. A key innovation is the development of the TJYRoad-Net model, which achieves over 85% mIoU segmentation accuracy by including a traffic feature computing (TFC) module composed of three critical components: the Regional Coordinate Encoder (RCE), the Context-Aware Aggregation Unit (CAU), and the Hierarchical Expansion Block. Comparative analysis segments the point clouds into road and non-road categories, achieving centimeter-level registration accuracy with RANSAC and ICP. Two lightweight surface reconstruction techniques are implemented: (1) algorithmic reconstruction, which delivers a 6.3 mm elevation error at 95% confidence in complex intersections, and (2) template matching, which replaces road markings, poles, and vegetation using bounding boxes. These methods ensure accurate results with minimal memory overhead. The optimized 3D models have been successfully applied in driving simulation and traffic flow analysis, providing a practical and scalable solution for real-world infrastructure modeling and analysis. These applications demonstrate the versatility and efficiency of the proposed methods in modern traffic system simulations.

Fault3DNnet: A lightweight 3D seismic fault detection network with bidirectional decoding

Fault detection from seismic data is a critical component of hydrocarbon exploration and production. In recent years, deep-learning techniques have made tremendous achievements and breakthroughs in seismic fault detection. Especially U-nets, represented by FaultSeg3D and its variants, are a dominant method for 3D deep-learning-based seismic fault interpretation. The incorporation of state-of-the-art network modules or training strategies with U-nets continuously improves the performance of fault detection for field seismic data, usually at the cost of increased computational intensity and hardware consumption. We design a lightweight bidirectional decoding network, Fault3DNnet, for fault detection from 3D seismic data volume. To make it simple and concise, our Fault3DNnet is equipped with basic conventional convolution and regular operation modules. The two decoding branches in Fault3DNnet could provide complementary information and improve fault detection performance. The superiority of Fault3DNnet is verified using publicly available synthetic and four field seismic data sets. The experimental results indicate that the method predicts fault structures with improved completeness and continuity compared with FaultSeg3D. Meanwhile, the parameter and computational quantity of the network are only 1/7 and 1/5 of FaultSeg3D, respectively. With the same graphics processing unit memory consumption, Fault3DNnet can handle the size of approximately six times the volume of 3D seismic data during inferencing as that of FaultSeg3D. It helps to improve prediction stability and decrease discontinuous artifacts at the stitching boundaries when predicting the large field data in chunks. In general, our Fault3DNnet delivers superior fault detection results while greatly reducing computational intensity and memory usage.

3D LVCN: A Lightweight Volumetric ConvNet

ABSTRACTIn recent years, with the significant increase in the volume of three‐dimensional medical image data, three‐dimensional medical models have emerged. However, existing methods often require a large number of model parameters to deal with complex medical datasets, leading to high model complexity and significant consumption of computational resources. In order to address these issues, this paper proposes a 3D Lightweight Volume Convolutional Neural Network (3D LVCN), aiming to achieve efficient and accurate volume segmentation. This network architecture combines the design principles of convolutional neural network modules and hierarchical transformers, using large convolutional kernels as the basic framework for feature extraction, while introducing 1 × 1 × 1 convolutional kernels for deep convolution. This improvement not only enhances the computational efficiency of the model but also improves its generalization ability. The pro‐posed model is tested on three challenging public datasets, namely spleen, liver, and lung, from the medical segmentation decathlon. Experimental results show that the proposed model performance has in‐creased from 0.8315 to 0.8673, with a reduction in parameters of approximately 5%. This indicates that compared to currently advanced model structures, our proposed model architecture exhibits significant advantages in segmentation performance.

Development of A Workshop on the Topic of the Bionically Inspired Lightweight Design Method for 3D Printed Components

Additive manufacturing methods, particularly 3D printing, are widely utilized in research and engineering for crafting lightweight yet durable materials capable of withstanding substantial forces. Leveraging insights from biomechanical structures offers a deeper understanding of reinforcement techniques and optimal design strategies to enhance strength. This paper focuses on the development of a Draisine design, historically significant in Karlsruhe, as part of the BWSplus project "Drais3D-Trinational" workshop. Through a comprehensive analysis of bionic influences on lightweight design and 3D printing parameters, this study aims to create an optimal design framework for the workshop. Utilizing Computer-Aided Design (CAD) software such as SolidWorks and Creo Parametric, along with Finite Element Analysis using ANSYS R2023 Workbench, deformation and stress analysis are conducted. Investigation into 3D printing parameters, including infill patterns, temperature, support systems, and orientation, seeks to identify optimal solutions considering factors like processing time, robustness, and filament wastage. The objective is to explore the biomechanical influence on construction methods, concept design, and parameter construction in preparation for the Drais3D-Trinational workshop in March 2023. The expected outcomes include the presentation of two main designs with varied parameters, alongside analyses such as Finite Element Analysis and optimization of 3D printing parameters, emphasizing the role of bionic structures in defining the optimal Draisine design.

Performance of normal-weight and lightweight 3D printed cementitious composites with recycled glass: Sorption and microstructural perspective

The development of sustainable mixtures has been a widely researched topic in the last decade, as the cement and concrete industries are among the most polluting sectors. Recycled materials can eliminate the need for extraction of natural materials through re-utilization of waste, reducing the environmental impact. To date, few studies have analyzed the effects of incorporating recycled glass aggregates on the water transport properties and microstructure of 3D printable mixtures.This study assesses the feasibility of incorporating recycled glass as a replacement for natural aggregate in 3D printable mixtures. For this purpose, three normal-weight and three lightweight mixtures containing 0, 50, and 100 vol.% of recycled glass aggregate as basalt aggregate replacement were evaluated in cast and printed samples. Expanded thermoplastic microspheres (ETM) were included to create the lightweight mixture composition. This experimental work evaluated the effects of incorporating recycled glass, ETM, and the impact of the printing process on the microstructure, sorption, and capillary water porosity (CWP).The printing process has proven to be an influential factor affecting the shape, size, and quantity of porosity. In addition, the printing process was demonstrated to be more influential on porosity formation than the inclusion of recycled glass or ETM. Incorporating recycled glass results in a slight decrease in the PHg porosity, more spherical pores, and lower anisotropic tendency. Furthermore, adding ETM has a filling effect, leading to a more compact microstructure.

A lightweight network-based sign language robot with facial mirroring and speech system

Human–robot interaction is an essential capability for humanoid robots to enter the physical world and become companions in people’s lives, learning, and work. While the majority of current research focuses on the voice-based interactions of robots, yet over 60% of communication occurs through nonverbal behaviors, such as facial expressions and hand gestures. Endowing robots with the ability to communicate through nonverbal behavior not only enhances the interactive experience with robots but also provides a potential communication tool for individuals with hearing or speech impairments. Here, we develop a humanoid robot capable of adjusting facial movements by driving servos, and design a novel framework for the robot to integrate sign language recognition and facial landmark detection algorithms. This framework facilitates the robot recognize sign language and translate it into spoken language, while also imitating the facial expressions of the signers. To achieve this, we also propose a lightweight deep learning network called RealTimeSignNet for real-time sign language recognition. Leveraging lightweight 3D convolution modules and time-dependent constraints, this model adapts to various time scales, ensuring efficient processing of sign language recognition tasks. Experimental results demonstrate the outstanding performance of the RealTimeSignNet model on mainstream sign language datasets, achieving an accuracy of 88.1% on the large continuous sign language dataset (continuous SLR), 98.2% on the isolated sign language dataset (SLR 500), and 91.50% on the English sign language dataset (WLAS). The overall assessment demonstrates that our humanoid robot is capable of recognizing sign language and translating it into spoken language, while imitating the facial emotions, providing a comprehensive solution to the communication challenges faced by individuals with hearing and speech impairments.

3D Lightweight Cryptosystem Design For IoT Applications Based On Composite S-Box

The security of the Internet of Things (IoT) depends on strong cryptographic functions that require extensive computation and resources. The security of Internet of Things (IoT) relies on the use of strong cryptographic functions that demand extensive computation and resources. Therefore, the selection of a cryptographic system is influenced by the computational and communicational capabilities of IoT devices, including their energy requirements, memory constraints, and execution time. This paper aims to develop a lightweight and secure encryption algorithm for IoT applications, focusing on advancing cryptanalysis techniques. We propose a new lightweight algorithm named enhanced 3D RECTANGLE, designed to deliver robust security for IoT applications and optimized for cell phones with minimal memory usage, low power consumption, and efficient performance. The RECTANGLE block cipher was chosen for this research due to its high efficiency and speed relative to other lightweight algorithms, despite some security issues. The proposed enhanced 3D RECTANGLE algorithm improves confusion and diffusion properties through a new 3D array block rotation method for 4x4 plaintext, based on a 128-bit key and 16 rounds. Additionally, we designed a 4x4 composite S-Box with a Galois field pipelining structure, offering a more advanced solution compared to the Look-Up Table method. The cryptanalysis, avalanche effect, and bit error rate tests were conducted to verify the security strength of the proposed algorithm. The proposed algorithm was evaluated against two existing algorithms, RECTANGLE and Extended 3D RECTANGLE. The enhanced 3D RECTANGLE algorithm provides better cryptanalysis, demonstrating a 40% avalanche effect and achieving a bit error rate (BER) of 31%, indicating its greater effectiveness compared to the other two lightweight algorithms.

Design, Experiments, and Path Planning for a Lightweight 3D Minimally Actuated Serial Robot with a Mobile Actuator

This paper presents a novel three-dimensional (3D) minimally actuated serial robot (MASR) and its unique kinematic analysis. Unlike traditional robots, the 3D MASR features a passive arm devoid of wires or motors, comprising passive rotational and prismatic joints. A single mobile actuator (MA) traverses the arm, engages designated joints for operation, and locks them in place with a worm gear setup. A gripper is attached to the MA, enabling object transportation along the arm, reducing joint actuation, and optimizing task completion time. Our key contributions include the mechanical design, and in particular the robot’s passive joints with their automated actuation mechanism, and a novel optimization algorithm leveraging neural networks (NNs) to minimize task completion time through advanced kinematic analysis. Experiments with a physical prototype of the 3D MASR demonstrate its major advantages: it is remarkably lightweight (2.3 kg for a 1 m long arm and a 1 kg payload) compared to similar robots; it is highly modular (five joints R and P actuated by a single MA); and part replacement is effortless due to the absence of wiring along the arm. These features are visually depicted in the accompanying video.

Easy to fabricate 3D metastructure for low-frequency vibration control

As a burgeoning category of elastic metamaterials, 3D metastructures have garnered significant research attention for manipulating low-frequency acoustic and elastic waves. Bandgap engineering allows for the control of these waves across a subwavelength ultrawide frequency range. However, the manufacturing of these 3D structures poses a challenge, necessitating additional support materials for 3D-printed components, creating difficulties in mass production. In this study, we propose a novel lightweight 3D metastructure design that is easy to fabricate and provides a low-frequency subwavelength bandgap. We replaced conventional struts supporting heavy mass inclusions in typical designs with modified arch beams. This structural modification enables the easy and self-supporting manufacturing of 3D metastructure unit cells without the need for extra support material. Utilizing magnets and steel masses with bolts as hard inclusions, the magnet facilitates the quick assembly of the 3D metastructure, potentially facilitating mass manufacturing in practical applications. The wave dispersion and bandgap properties of the metastructure are investigated numerically, and experimental vibration tests are performed on the 3D-printed and assembled parts. The experimental results and numerical findings demonstrate robust vibration attenuation at low frequencies by the proposed 3D metastructure. The suggested, easy-to-fabricate 3D-metastructure design holds potential applications in low-frequency elastic-wave manipulation, including noise and vibration control.

Curvature-induced strain to realize differential lithiophilicity for selective lithium deposition and stable lithium anode

Regulating the lithiophilicity of three-dimensional (3D) lithium hosts is crucial for improving lithium deposition and suppressing dendrite growth. However, previous research has mainly focused on varying lithiophilic components, and the impact of geometric structures induced strain remains poorly understood. Herein, we present a novel curvature-induced strain engineering approach to regulate lithiophilicity and deposition kinetics of lithium in lightweight 3D tubular carbon hosts. A hollow carbon fiber with bilateral growth of MnO2 nanosheet arrays both inside and outside the tubes (MnO2@HCFC) have been successfully synthesized. Theoretical calculations and experiments confirm that the MnO2 layers inside and outside the hollow carbon fiber tube undergo compressive and tensile strain, respectively. The curvatures induced strain modifies the O2p band center of the lithiophilic MnO2 layer, enabling the regulation of lithiophilicity and preferential and uniform lithium deposition within the carbon fiber tubes. The MnO2@HCFC framework exhibits excellent lithium affinity and uniform Li+ flux distribution, as evidenced by visualization techniques and COMSOL simulations, enabling dendrite-free lithium deposition. The Li-MnO2@HCFC||LiCoO2 full cell retains 85.7 % of its capacity after 400 cycles at 0.5 C with a high LiCoO2 loading of 10.3 mg cm-2. The optimized lithium anode pouch cell exhibits robust cycle stability under harsh conditions. This work offers new insight into the design of 3D lithium hosts to enhance the performance of lithium anodes through curvature-induced strain engineering.

ICNoduleNet: Enhancing Pulmonary Nodule Detection Performance on Sharp Kernel CT Imaging.

Thoracic computed tomography (CT) currently plays the primary role in pulmonary nodule detection, where the reconstruction kernel significantly impacts performance in computer-aided pulmonary nodule detectors. The issue of kernel selection affecting performance has been overlooked in pulmonary nodule detection. This paper first introduces a novel pulmonary nodule detection dataset named Reconstruction Kernel Imaging for Pulmonary Nodule Detection (RKPN) for quantifying algorithm differences between the two imaging types. The dataset contains pairs of images taken from the same patient on the same date, featuring both smooth (B31f) and sharp kernel (B60f) reconstructions. All other imaging parameters and pulmonary nodule labels remain entirely consistent across these pairs. Extensive quantification reveals mainstream detectors perform better on smooth kernel imaging than on sharp kernel imaging. To address suboptimal detection on the sharp kernel imaging, we further propose an image conversion-based pulmonary nodule detector called ICNoduleNet. A lightweight 3D slice-channel converter (LSCC) module is introduced to convert sharp kernel images into smooth kernel images, which can sufficiently learn inter-slice and inter-channel feature information while avoiding introducing excessive parameters. We conduct thorough experiments that validate the effectiveness of ICNoduleNet, it takes sharp kernel images as input and can achieve comparable or even superior detection performance to the baseline that uses the smooth kernel images. The evaluation shows promising results and proves the effectiveness of ICNoduleNet.

Investigation of the static and low‐velocity impact performance of 3D braided spacer composites

AbstractTo meet the lightweight requirements of advanced composite materials, the ultralightweight three‐dimensional (3D) braided spacer composites were designed and manufactured by ultra‐high molecular weight polyethylene (UHMWPE). The key structural effects of core column distances and core column heights on the static bending, compression, and low‐velocity impact properties of the composites were further investigated. Results indicated that the mechanical properties of 3D braided spacer composites was closely depended on their structural parameters. The flexure and compression of 3D braided spacer composites decreased with increasing core column distance and core column height. W1/H1 specimen showed the optimal bending and compressive performance. The maximum flexural strength and flexural modulus were approximately 54.77 MPa and 4.95 GPa, respectively. The low‐velocity impact properties were further investigated where the stiffness, energy absorption, and impact resistance of the composite decreased with increasing core column distance. The peak load and energy absorption of the W1 specimen were 2.39 kN and 7.99 J, respectively. While the low‐velocity impact properties increased with increasing core column distance. The H3 specimen exhibited the greatest stiffness, impact resistance and energy absorption with peak loads and energy absorption of 2.81 kN and 8.49 J, respectively. In addition, the macro morphology and scanning electron microscopy (SEM)micrographs were examined to investigate the damage morphology and failure mechanism of the composites. The main failure modes were matrix cracking, interface debonding, fiber buckling, and dislocation, as well as tilting, and failure of the core columns. This work provides guidance for structural design and engineering application of 3D lightweight braided spacer composite.

Adjusting Ion Diffusion Kinetics of Li Deposition Enabled by an Elastic Porous Melamine Sponge Host for Stable Lithium Metal Anodes.

Lithium (Li) dendritic growth and huge volume expansion seriously hamper Li-metal anode development. Herein, we design a lightweight 3D Li-ion-affinity host enabled by silver (Ag) nanoparticles fully decorating a porous melamine sponge (Ag@PMS) for dendrite-free and high-areal-capacity Li anodes. The compact Ag nanoparticles provide abundant preferred nucleation sites and give the host strong conductivity. Moreover, the high specific surface area and polar groups of the elastic, porous melamine sponge enhance the Li-ion diffusion kinetics, prompting homogeneity of Li deposition and stripping. As expected, the integrated 3D Ag@PMS-Li anode delivered a remarkable electrochemical performance, with a Coulombic efficiency (CE) of 97.14% after 450 cycles at 1 mA cm-2. The symmetric cell showed an ultralong lifespan of 3400 h at 1 mA cm-2 for 1 mAh cm-2. This study provides a facile and cost-effective strategy to design an advanced 3D framework for the preparation of a stable dendrite-free Li metal anode.

MFE‐MVSNet: Multi‐scale feature enhancement multi‐view stereo with bi‐directional connections

AbstractRecent advancements in deep learning have significantly improved performance in the multi‐view stereo (MVS) domain, yet achieving a balance between reconstruction efficiency and quality remains challenging for learning‐based MVS methods. To address this, we introduce MFE‐MVSNet, designed for more effective and precise depth estimation. Our model incorporates a pyramid feature extraction network, featuring efficient multi‐scale attention and multi‐scale feature enhancement modules. These components capture pixel‐level pairwise relationships and semantic features with long‐range contextual information, enhancing feature representation. Additionally, we propose a lightweight 3D UNet regularization network based on depthwise separable convolutions to reduce computational costs. This network employs bi‐directional skip connections, establishing a fluid relationship between encoders and decoders and enabling cyclic reuse of building blocks without adding learnable parameters. By integrating these methods, MFE‐MVSNet effectively balances reconstruction quality and efficiency. Extensive qualitative and quantitative experiments on the DTU dataset validate our model's competitiveness, demonstrating approximately 33% and 12% relative improvements in overall score compared to MVSNet and CasMVSNet, respectively. Compared to other MVS networks, our approach more effectively balances reconstruction quality with efficiency.

Virtual Reality Scenario Analysis of Art Design Taking into Account Interactive Digital Media Pattern Generation Technology

In art design, 3D printing technology is crucial, and more and more creators conceive scenes using 3D modeling software to get a three-dimensional and beautiful work. Due to the large amount of noise and redundant points in the raw data collected during the modeling process, the generation speed and rendering effect of 3D models are reduced. Given the above problems, the study designed an interactive 3D lightweight modeling system based on the combination of hand-drawn sketching and laser 3D scanning based on the streamlined algorithm. The experimental results showed that when the hand-drawn speed was 300, the number of triangular slices, model size, and time required to generate the model of the hand-drawn sketching model based on the streamlined algorithm were reduced by 67.39%, 65.48%, and 63.79%, respectively. In the real-time point cloud data streamlining process of the laser 3D scanning model, the point cloud data reduction ratio and the streamlining goodness index of the point cloud streamlining algorithm are 71.99% and 3.06%, respectively. The system performance is robust, and the data processing speed and rendering effect are good.

Automatic liver segmentation from CT volumes based on multi-view information fusion and condition random fields

Liver segmentation from CT images is a fundamental and essential step for many clinical applications, such as disease diagnosis, surgery navigator, and radiation therapy. Due to the limited computing resources, automatic and accurate liver segmentation from 3D CT volume is still a challenge task. In this paper, we propose an automatic liver segmentation method by fusing multi-view information. First, a 2D network combined with Dual Self-Attention (DSA) is proposed and applied on multi-sectional 2D slices reconstructed from axial, coronal, and sagittal directions. Then, to generate 3D segmentation results, a lightweight 3D network is designed to fuse the segmentation results from different viewing. Finally, a full connection condition random field is used to refine the 3D results. The proposed method is evaluated on four public datasets, including 3DIRCADb, CHAOS, Sliver07, and LiTS. The experimental results show that the proposed method can accurately segment the liver region for the datasets with large or small cases, as well as healthy or lesion livers. The dice values obtained by the proposed method on four datasets are 0.952, 0.969, 0.958, and 0.964, respectively, which are better than those obtained by many existing ones.

Topology optimization of forming tools: pressure die in rotary draw bending process

AbstractThe forming tools are conventionally manufactured from alloy steels. These tools can be designed to be light weight and cost effective without compromising performance characteristics. In this research, a fundamental forming tool (pressure die) of ‘rotary draw bending processes’ is topology optimized and 3D printed by using polymeric material. An accurate FE-simulation model of a rotary draw bending process is developed and topology optimization of the pressure die is carried out on the basis of contact stresses provided by FE-simulation results. The topology optimized pressure die is compared with its conventional metal made counter-part in terms of contact forces, contact normal stresses, effective area in-contact and cost performance index. This manuscript demonstrates that topology optimized lightweight 3D printed polymeric forming tools are a cost effective alternative to comparatively heavy alloy steels. This research contributes to widen the avenue of cost effective manufacturing by incorporating topology optimization regimes into existing production setups.

Simulation and experimental study on rope driven artificial hand and driven motor.

Prosthetic hands have the potential to replace human hands. Using prosthetic hands can help patients with hand loss to complete the necessary daily living actions. This paper studies the design of a bionic, compact, low-cost, and lightweight 3D printing humanoid hand. The five fingers are underactuated, with a total of 9 degrees of freedom. In the design of an underactuated hand, it is a basic element composed of an actuator, spring, rope, and guide system. A single actuator is providing power for five fingers. And the dynamic simulation is carried out to calculate the motion trajectory effect. In this paper, the driving structure of the ultrasonic motor was designed, and the structural size of the ultrasonic motor vibrator was determined by modal and transient simulation analysis, which replace the traditional brake, realize the lightweight design of prosthetic hand, improve the motion accuracy and optimize the driving performance of prosthetic hand. By replacing traditional actuators with new types of actuators, lightweight design of prosthetic hands can be achieved, improving motion accuracy and optimizing the driving performance of prosthetic hands.

Organ Segmentation and Phenotypic Trait Extraction of Cotton Seedling Point Clouds Based on a 3D Lightweight Network

Cotton is an important economic crop; therefore, enhancing cotton yield and cultivating superior varieties are key research priorities. The seedling stage, a critical phase in cotton production, significantly influences the subsequent growth and yield of the crop. Therefore, breeding experts often choose to measure phenotypic parameters during this period to make breeding decisions. Traditional methods of phenotypic parameter measurement require manual processes, which are not only tedious and inefficient but can also damage the plants. To effectively, rapidly, and accurately extract three-dimensional phenotypic parameters of cotton seedlings, precise segmentation of phenotypic organs must first be achieved. This paper proposes a neural network-based segmentation algorithm for cotton seedling organs, which, compared to the average precision of 75.4% in traditional unsupervised learning, achieves an average precision of 96.67%, demonstrating excellent segmentation performance. The segmented leaf and stem point clouds are used for the calculation of phenotypic parameters such as stem length, leaf length, leaf width, and leaf area. Comparisons with actual measurements yield coefficients of determination R2 of 91.97%, 90.97%, 92.72%, and 95.44%, respectively. The results indicate that the algorithm proposed in this paper can achieve precise segmentation of stem and leaf organs, and can efficiently and accurately extract three-dimensional phenotypic structural information of cotton seedlings. In summary, this study not only made significant progress in the precise segmentation of cotton seedling organs and the extraction of three-dimensional phenotypic structural information, but the algorithm also demonstrates strong applicability to different varieties of cotton seedlings. This provides new perspectives and methods for plant researchers and breeding experts, contributing to the advancement of the plant phenotypic computation field and bringing new breakthroughs and opportunities to the field of plant science research.

Automatic Evaluation Method for Functional Movement Screening Based on Multi-Scale Lightweight 3D Convolution and an Encoder–Decoder

Functional Movement Screening (FMS) is a test used to evaluate fundamental movement patterns in the human body and identify functional limitations. However, the challenge of carrying out an automated assessment of FMS is that complex human movements are difficult to model accurately and efficiently. To address this challenge, this paper proposes an automatic evaluation method for FMS based on a multi-scale lightweight 3D convolution encoder–decoder (ML3D-ED) architecture. This method adopts a self-built multi-scale lightweight 3D convolution architecture to extract features from videos. The extracted features are then processed using an encoder–decoder architecture and probabilistic integration technique to effectively predict the final score distribution. This architecture, compared with the traditional Two-Stream Inflated 3D ConvNet (I3D) network, offers a better performance and accuracy in capturing advanced human movement features in temporal and spatial dimensions. Specifically, the ML3D-ED backbone network reduces the number of parameters by 59.5% and the computational cost by 77.7% when compared to I3D. Experiments have shown that ML3D-ED achieves an accuracy of 93.33% on public datasets, demonstrating an improvement of approximately 9% over the best existing method. This outcome demonstrates the effectiveness of and advancements made by the ML3D-ED architecture and probabilistic integration technique in extracting advanced human movement features and evaluating functional movements.