Variation In Shape Research Articles

Degradation Detection in Rice Products via Shape Variations in XCT Simulation-Empowered AI

This research explores the process of generating artificial training data for the detection and classification of defective areas in X-ray computed tomography (XCT) scans in the agricultural domain using AI techniques. It aims to determine the minimum detectability limit for such defects through analyses regarding the Probability of Detection based on analytic XCT simulations. For this purpose, the presented methodology introduces randomized shape variations in surface models used as descriptors for specimens in XCT simulations for generating virtual XCT data. Specifically, the agricultural sector is targeted in this work in terms of analyzing common degradation or defective areas in rice products. This is of special interest due to the huge biological genotypic and phenotypic variations occurring in nature. The proposed method is demonstrated on the application of analyzing rice grains for common defects (chalky and pore areas).

Morphological assessment of mandible: Way to effective anaesthesia

The biggest bone in the human skull, the mandible shapes the inferior third of the face and forms the lower jawline. The inferior alveolar nerve and vessels enter the mandibular canal through the mandibular foramen, which punctures the internal surface of the ramus. It is located at the level of the occlusal surfaces of the lower teeth, midway between the anterior and posterior borders of the ramus, and ends at the mental foramen. Most people have a single mandibular canal, while there are some variations in shape (oval, round, or pear-shaped) and whether an accessory canal can be distinguished (canal bifurcation). Failure of inferior alveolar nerve block (IANB) anesthesia can be attributed to these anatomical variables as well as operator technique. The anesthetic solution is administered as near to the mandibular foramen as feasible by adhering to the landmarks. Repeated injections of the local anesthetic solution in children due to IANB failure can be a tiresome task since they may cause the kid to behave negatively and there is a chance that the solution will be administered in excess of the safe suggested dosage. Data regarding the shifting position of the mandibular foramen from childhood to puberty is lacking in the literature. In light of this, this research was carried out to evaluate, using radiographic techniques, how the mandibular foramen's locations change as a kid grows and develops from childhood to adolescent.

ASFormer: Attentive Semantic Feature Fusion Transformer for Pixel -level Defect Detection

Abstract Surface defect detection is pivotal for ensuring product quality in various industries. These defects typically manifest as low background contrast, substantial variations in shape, and a scarcity of balanced training samples. Traditional defect detection methods often encounter limitations in terms of low detection accuracy and insufficient detection granularity. This study introduces an innovative attentive semantic feature fusion transformer network (ASFormer) tailored for pixel-level defect detection tasks to overcome these challenges. Initially, the transformer architecture is employed to extract the multiscale features of defects. Subsequently, a multiscale semantic fusion module is integrated, complemented by a dynamic upsampling mechanism, to mitigate the loss of detailed information during feature fusion. Moreover, a scale-aware dual-attention module is developed, which effectively captures the intricacies of the fused features across both channel and spatial dimensions, ensuring pixel-level detection precision. Additionally, a contextual boundary loss function is proposed to augment the ability of the network to discern defect boundaries, different categories and defect scales. Experimental validation on the NEU-Seg and Crack500 surface defect datasets demonstrated that the proposed ASFormer achieved state-of-the-art performance, with mIoU scores of 85.24% and 78.20%, respectively.

Accurate retinal vessel segmentation in OCT-A images based on MsTCG-Net

Optical coherence tomography angiography (OCT-A) is a non-invasive visualization imaging technology with high-resolution that can more clearly image tiny blood vessels. Using OCT-A imaging technology, certain ophthalmic diseases can be better diagnosed by the morphological changes of retinal blood vessels. However, the task of segmenting retinal vessels is still very challenging due to the large variation in vessel size and shape and the presence of noise. In this paper, by introducing a transformer with a self-attention mechanism, we propose a novel multi-scale transformer-based channel and global attention network (MsTCG-Net) for segmentation of blood vessels in retinal OCT-A images. In MsTCG-Net, transformer-based channel joint attention (TC) block and transformer-based global joint attention (TG) block are proposed to capture multi-semantic features from spatial and channel dimension and fuse global contextual semantic features from different layers of encoder. Experimental results show that our proposed method achieves better segmentation performance than other state-of-the-art U-Net-based methods.

Accurate retinal vessel segmentation in OCT-A images based on MsTCG-Net

Optical coherence tomography angiography (OCT-A) is a non-invasive visualization imaging technology with high-resolution that can more clearly image tiny blood vessels. Using OCT-A imaging technology, certain ophthalmic diseases can be better diagnosed by the morphological changes of retinal blood vessels. However, the task of segmenting retinal vessels is still very challenging due to the large variation in vessel size and shape and the presence of noise. In this paper, by introducing a transformer with a self-attention mechanism, we propose a novel multi-scale transformer-based channel and global attention network (MsTCG-Net) for segmentation of blood vessels in retinal OCT-A images. In MsTCG-Net, transformer-based channel joint attention (TC) block and transformer-based global joint attention (TG) block are proposed to capture multi-semantic features from spatial and channel dimension and fuse global contextual semantic features from different layers of encoder. Experimental results show that our proposed method achieves better segmentation performance than other state-of-the-art U-Net-based methods.

Waterways and Roads in the Pre-colonial Bengal Delta: A Historical Perspective

The paper reviews the general conditions of the transport networks in the pre-colonial Bengal Delta. This region has a rich history of political dynamics, a thriving economy, and active commercial ventures. In the pre-colonial period, two primary transport systems operated in the Bengal Delta: waterways, due to the riverine and coastal landscape, served as the main mode of communication, while road transport was utilised in the areas where water transport was minimal. Variations in nature, capacity, shape, and functionality of these transport systems were influenced by the distinct geo-environmental features of the region. This paper serves as an in-depth exploration of pre-colonial transport systems, providing a foundation for understanding the evolution of modern transportation in Bengal.

Damage assessments of designed sandwich cores against an idealized impact loading of ship accidents: experiment-based validated nonlinear finite element approach

PurposeThe high rate of ship casualties demands serious attention. Earlier studies have yet to integrate several scenarios in an experiment. This research aims to determine damage to a ship's hull that is assumed to have been stranded due to being hit by a rock under the sea.Design/methodology/approachThis scenario was analyzed by conducting a penetration test simulation using the ANSYS LS DYNA. Modeling was carried out with core, speed, indenter shape and indenter angle variations. The test is carried out by moving the indenter against the plate until the plate is damaged.FindingsThe results of this research show that the round indenter gave the best results in testing, while changes in speed below 1.5 m/s did not significantly affect the results. The contact angle of the indenter to the panel is directly proportional to the damage to the panel, and core X has the best results in the test.Originality/valueThis work presented several proposed eccentric sandwich panels to calculate their performance against impact loading. The detailed geometry for future reference is presented, while the indenter-panel interactions are validated based on experimental work.

Solar shape variations across cycles 24 and 25: Observations from 2010 to 2023

The longest continuous time-series of solar oblateness measurements, initiated in 2010 and still ongoing, has been obtained from data collected by the Helioseismic Magnetic Imager (HMI) aboard NASA's Solar Dynamics Observatory (SDO). Based on HMI data, we developed two methods for determining the solar oblateness at 617.33\,nm in the continuum . The first method involves determining solar oblateness using HMI solar disk images and limb observations from twenty-three SDO satellite roll calibration maneuvers between 2010 and 2023 . Through meticulous analysis of these observation sequences, we obtained a precise measurement of solar oblateness using this technique, yielding a value of 9.02 (pm 0.72)times 10$^ $ (6.28pm 0.50\,kilometers), unaffected by brightness contamination from sunspots and magnetically induced excess emission. We also verified the polarization independence of light, showing consistent HMI solar oblateness measurements across Stokes states. Interestingly, our solar oblateness time-series, based on HMI solar disk images and limb observations, seems to be in anti-phase with solar activity. The second method we used relies on determining solar oblateness from HMI helioseismic inference of internal rotation. With this approach, we obtained a solar oblateness of 8.40 (pm 0.02)times 10$^ $ (5.85pm 0.01 kilometers) with a variation in phase with solar activity (0.05times 10$^ $ (0.04\,kilometers at 1\,sigma ) over an 11--year sunspot cycle). This outcome is troubling as it conflicts with our results obtained from the HMI solar limb observations

Wide Variation in Shape of Hypoplastic Left Ventricles Undergoing Recruitment and Biventricular Repair: A Statistical Shape Modeling Study.

Patients with hypoplastic left ventricles (LV) who undergo volume-loading procedures (recruitment, biventricular (BIV) repair) are at risk for adverse outcomes including heart failure and death. We investigated pre-BIV LV shape as a predictor of outcome after BIV repair in patients with hypoplastic LVs. Baseline and post-recruitment cardiac MRI and CT data were analyzed in patients with hypoplastic LV (< 50 ml/m2). Statistical shape modeling (SSM) was utilized to generate a model of the shape and variability of LVs. Traditional measures of LV sphericity and eccentricity were also measured. Major adverse cardiac events (MACE) included heart failure, transplant, and death. Of 95 patients with baseline mean LV volume 29 ± 13ml/m2, 45 (47%) had a right dominant atrioventricular canal defect, 31 (34%) had a variant of hypoplastic left heart syndrome, and 18 (19%) had endocardial fibroelastosis (EFE). A wide variation in LV shape was found by SSM, and shape modes were associated with RV and LV size, and diagnosis. BIV repair was achieved in 74 (78%) patients; and 13 (18%) of BIV patients had MACE. Predictors of MACE following BIV repair included EFE, higher RV mass index, and higher RV end-diastolic volume index. No baseline or post-recruitment LV shape parameter was associated with outcome after BIV repair. The shape model of hypoplastic LVs demonstrated a wide array of LV shapes. LVs gained sphericity and size and lost eccentricity with recruitment. Though the ventricles changed shape with recruitment, no specific LV shape characteristic at the baseline or post-recruitment stage was predictive of decision to proceed with BIV repair nor outcome. Higher RV mass and volume may represent new biomarkers that predict outcome following BIV repair in patients with hypoplastic LV. Further investigation could determine the reproducibility of these findings.

PSR-LeafNet: A Deep Learning Framework for Identifying Medicinal Plant Leaves Using Support Vector Machines

In computer vision, recognizing plant pictures has emerged as a multidisciplinary area of interest. In the last several years, much research has been conducted to determine the type of plant in each image automatically. The challenges in identifying the medicinal plants are due to the changes in the effects of image light, stance, and orientation. Further, it is difficult to identify the medicinal plants due to factors like variations in leaf shape with age and changing leaf color in response to varying weather conditions. The proposed work uses machine learning techniques and deep neural networks to choose appropriate leaf features to determine if the leaf is a medicinal or non-medicinal plant. This study presents a neural network design based on PSR-LeafNet (PSR-LN). PSR-LeafNet is a single network that combines the P-Net, S-Net, and R-Net, all intended for leaf feature extraction using the minimum redundancy maximum relevance (MRMR) approach. The PSR-LN helps obtain the shape features, color features, venation of the leaf, and textural features. A support vector machine (SVM) is applied to the output achieved from the PSR network, which helps classify the name of the plant. The model design is named PSR-LN-SVM. The advantage of the designed model is that it suits more considerable dataset processing and provides better results than traditional neural network models. The methodology utilized in the work achieves an accuracy of 97.12% for the MalayaKew dataset, 98.10% for the IMP dataset, and 95.88% for the Flavia dataset. The proposed models surpass all the existing models, having an improvement in accuracy. These outcomes demonstrate that the suggested method is successful in accurately recognizing the leaves of medicinal plants, paving the way for more advanced uses in plant taxonomy and medicine.

Development of ITER first wall heat load feedback control

Real-time monitoring and control of thermal loads on tungsten plasma-facing components is mandatory for ITER to avoid their overheating during plasma discharges. For the Start of Research Operations (SRO) phase, dedicated First Wall Heat Load Control (FWHLC) functions are included in the ITER Plasma Control System (PCS) to prevent the development of excessive plasma heat loads on the inertially cooled first wall (FW). In this work, we propose a FWHLC design with a model-based approach, which features a simplified thermal model for the FW armour. This control-oriented model simulates the thermal response of ITER FW to slow changes in plasma equilibrium and energy, which during ITER operation will be monitored by the real-time wide angle infrared camera system. This allows computationally efficient runs for rapid controller performance assessment but can also assist operation scenario design by pre-empting potential heat load issues. FWHLC is designed to react to measured and predicted FW temperatures overcoming predesigned limits with real-time variations of the plasma shape or auxiliary power references, aiming to reduce thermal loads before a premature plasma termination is required to protect the FW. The new scheme is tested in closed loop simulations, where perturbations to nominal ITER low current scenarios are introduced in the wall thermal model to trigger the response of FWHLC, supporting the effectiveness of the proposed policy.

Performance and geocell-soil interaction of sand subgrade reinforced with high-density polyethylene, polyester, and polymer-blend geocells: 3D numerical studies

Geocells are widely used in geotechnical and road applications due to their excellent reinforcing properties. However, variations in materials, joint configurations, pocket shapes, and other factors have led to geocell development in a wide range of types. Given this, the present study compared three typical geocells, polymer-blend geocells (PBGs), high-density polyethylene geocells (HDPEGs), and polyester (PETGs) geocells, in terms of pressure-settlement responses and geocell-soil interactions. A numerical model was firstly validated based on a plate loading test, and subsequently extended to analyze the sand subgrade reinforced with three geocell types. The numerical simulations considered geocells with varying weld distances, heights, and pocket-opening shapes. Numerical results showed that the reinforced performance of PETGs (diamond shape) was less sensitive to changes in wall height compared to PBGs and HDPEGs (honeycomb shape). There was little difference in the capacity to reduce soil bed settlement among the three geocell types, unless significant deformation occurs and the soil enters the plastic state. Furthermore, based on the shape of the opening pockets, the three types of geocells were classified into honeycomb-shaped and diamond-shaped categories to investigate geocell-soil interactions. The results of coupling shear stress, confining stress, and normal deformations demonstrated that diamond shapes may generate unreasonable lateral confining resistance, which was the main factor causing the geocell-soil interaction variation between these geocells.

The radiative torque spin-up efficiency of ballistic dust-grain aggregates

It is quintessential for the analysis of the observed dust polarization signal to understand the rotational dynamics of interstellar dust grains. Additionally, high rotation velocities may rotationally disrupt the grains, which impacts the grain-size distribution. We aim to constrain the set of parameters for an accurate description of the rotational spin-up process of ballistic dust grain aggregates driven by radiative torques (RATs). We modeled the dust grains as complex fractal aggregates grown by the ballistic aggregation of uniform spherical particles (monomers) of different sizes. A broad variation of dust materials, shapes, and sizes were studied in the presence of different radiation sources. We find that the canonical parameterization for the torque efficiency overestimates the maximum angular velocity $ RAT $ caused by RATs acting on ballistic grain aggregates. To resolve this problem, we propose a new parameterization that predicts $ RAT $ more accurately. We find that RATs are most efficient for larger grains with a lower monomer density. This manifests itself as a size- and monomer-density dependence in the constant part of the parameterization. Following the constant part, the parameterization has two power laws with different slopes that retain universality for all grain sizes. The maximum grain rotation does not scale linearly with radiation strength because different drag mechanisms dominate, depending on the grain material and environment. The angular velocity $ RAT $ of individual single dust grains has a wide distribution and may even differ from the mean by up to two orders of magnitude. Even though ballistic aggregates have a lower RAT efficiency, strong sources of radiation (stronger than $ 100$ times the typical interstellar radiation field) may still produce rotation velocities high enough to cause the rotational disruption of dust grains.

Text-independent speaker identification using modified SincNet with robust features from suitable acoustic region and appropriate optimizer for raw audio analysis

Speaker identification is a method of identifying an individual from a set of speakers, and text-independent speaker identification systems allow speakers to utter any phrase without any constraints. This study is focused on raw audio analysis as phase, fine-grained frequency patterns, timing cues, and other minute characteristics are preserved when raw waveforms are processed as compared to handcrafted features like Mel-Frequency Cepstral Coefficients (MFCC) and visual representation of audio-like spectrogram. Due to the depth of information, which includes variations in speech rhythm, pitch, and vocal tract shape, it is beneficial for identifying speakers. The deep learning architecture known as SincNet has gained popularity in speaker identification because of its parametric Sinc functions that allow it to operate directly on the raw audio input. In this paper, we have considered SincNet as the baseline model for speaker identification. The effect of proper speech boundary detection, including high-level features and effective optimizer selection are analysed. The precise identification of the signal start and terminus point is important for eliminating the redundant non-speech regions. We have included endpoint detection module as a pre-processing step in the system. Proper feature extraction and selection are crucial to the model's success. To extract more abstract features from the data, we have added more convolution layers to the original SincNet model. Further, we investigated the hyperparameter tuning protocol's sensitivity to the optimizer and selected the suitable optimizer for raw audio analysis. With all the modifications in the system architecture, we are able to archive improvements of 12.76 %, 13.33 %, and 13.39 % respectively for training, validation, and testing over the original SincNet model. In terms of validation loss, our proposed approach attains 0.35 in comparison to the original SincNet loss of 1.02. With this significant improvement, the total training time is marginally increased by 20 minutes for our proposed model. We have performed our investigation on the LibriSpeech dataset to check the effectiveness of our proposed system in comparison to the other model..

Variation of Stomach Shapes in Gastric Emptying Scintigraphy: Correlation With Gastric Emptying Results, Body Weight, and Symptoms.

Although different gastric shapes are encountered in gastric emptying scintigraphy (GES), it is not known whether gastric shape is related to gastric emptying (GE) or symptoms. To investigate different stomach shapes observed during GES and examine their associations with GE, body weight, and gastrointestinal (GI) symptoms. This was a retrospective review of GES studies performed at our institution. Patients with prior gastric surgery were excluded. A classification of gastric shapes included: crescentic, J-shaped, reversed-L, cylindrical, and bag-like. Gastric shapes were correlated with GE, BMI, and GI symptoms using PAGI-SYM. 397 GES studies were reviewed (317 females, age 40.0 ± 20.3 years, BMI 27.8 ± 13.8 kg/m2). Gastric shapes were: 41.8% crescentic/comma-shaped, 34.0% J-shaped, 19.9% reversed L-shape, 2.8% cylindrical, and 1.5% bag-like. BMI was highest in crescentic/comma-shaped stomachs (30.1 ± 12.4 kg/m2) and lowest in bag-like stomachs (25.2 ± 9.9 kg/m2; p = 0.022). Delayed GE was most pronounced in bag-like stomachs (34.4 ± 33.1% retention at 4 h) and lowest in reversed-L shape (10.3 ± 15.6% retention at 4 h; p = 0.008). Regurgitation severity was greatest in bag-like stomachs (2.2 ± 1.5) compared to milder symptoms in reversed-L shape (1.3 ± 1.4; p = 0.029). Heartburn severity was increased in bag-like stomachs (2.3 ± 1.6) while patients with cylindrical stomachs reported least heartburn severity (1.1 ± 1.3; p = 0.11). A classification system based on five gastric shapes observed during GES showed that crescent-shaped stomach was the most common shape and correlated with higher BMIs. Delayed GE was most pronounced in bag-like stomachs and lowest in reversed-L stomachs. Regurgitation and heartburn severity were greatest in bag-like stomachs with milder symptoms in reversed-L stomachs. Thus, gastric shape during GES is associated with gastric emptying, BMI, and symptom severity.

Disentangling Hippocampal Shape Variations: A Study of Neurological Disorders Using Mesh Variational Autoencoder with Contrastive Learning

This paper presents a comprehensive study focused on disentangling hippocampal shape variations from diffusion tensor imaging (DTI) datasets within the context of neurological disorders. Leveraging a Mesh Variational Autoencoder (VAE) enhanced with Supervised Contrastive Learning, our approach aims to improve interpretability by disentangling two distinct latent variables corresponding to age and the presence of diseases. In our ablation study, we investigate a range of VAE architectures and contrastive loss functions, showcasing the enhanced disentanglement capabilities of our approach. This evaluation uses synthetic 3D torus mesh data and real 3D hippocampal mesh datasets derived from the DTI hippocampal dataset. Our supervised disentanglement model outperforms several state-of-the-art (SOTA) methods like attribute and guided VAEs in terms of disentanglement scores. Our model distinguishes between age groups and disease status in patients with Multiple Sclerosis (MS) using the hippocampus data. Our Mesh VAE with Supervised Contrastive Learning shows the volume changes of the hippocampus of MS populations at different ages, and the result is consistent with the current neuroimaging literature. This research provides valuable insights into the relationship between neurological disorder and hippocampal shape changes in different age groups of MS populations using a Mesh VAE with Supervised Contrastive loss.&lt;br&gt;Our code is available at &lt;a href='https://github.com/Jakaria08/Explaining_Shape_Variability'&gt;https://github.com/Jakaria08/Explaining_Shape_Variability&lt;/a&gt;

Local and Regional, Not Latitudinal, Variation in Microclimate and Bedrock Shapes Moss‐Associated Diatom Communities in Greenland

ABSTRACTAimGreenland is among the most rapidly changing regions on Earth, with climate change having a profound impact on its terrestrial ecosystems. Here, moss‐associated diatoms hold great potential as sensitive biological indicators to monitor responses to climate change, but their diversity, community structure and biogeography remain virtually unexplored. Our study, thus, aims to (1) explore the diversity and community structure of moss‐associated diatoms in Greenland and (2) assess the environmental and spatial variables driving their geographic distribution, establishing a baseline for their use as bio‐indicators of climate change.LocationGreenland, spanning a gradient from high to subarctic regions.TaxonMoss‐associated diatoms (Bacillariophyta).MethodsLM and SEM analysis was conducted on 175 terrestrial moss samples collected from preserved herbarium material and fresh samples (1988–2021) from high, low and subarctic localities in Greenland. Biogeographical distributions, biodiversity patterns and community structure were examined in relation to environmental and spatial factors using Kruskal–Wallis, Spearman's Rank, ordination, SIMPROF and variation partitioning analyses.ResultsA total of 544 diatom taxa (66 genera) were identified, with nearly half potentially new to science. Community structure was predominantly influenced by moisture, pH, conductivity and temperature. Local variation in microclimate and bedrock, thus, explained more variation in the diatom communities than latitude‐related environmental gradients. Surprisingly, temperature had a negative impact on diatom richness, probably due to its adverse effect on moisture, highlighting the vulnerability of Arctic moss diatoms to global warming.Main ConclusionsThis study provides the first comprehensive analysis of Greenland's moss‐associated diatoms, highlighting their high diversity and sensitivity to environmental changes. Our findings underscore their potential as indicators for monitoring climate change in the Arctic, with moisture, pH, conductivity and temperature being critical factors influencing their communities, laying the groundwork for future research and monitoring efforts.

Note on the intraspecific variation of style shape, a trait of potential taxonomic value, in pea (Pisum sativum)

Note on the intraspecific variation of style shape, a trait of potential taxonomic value, in pea (Pisum sativum)

A framework for Interpretable deep learning in cross-subject detection of event-related potentials

Event-related potential-based Brain-Computer Interfaces are becoming widely popular due to their ability to send commands quickly with high accuracy. However, the stationary characteristics of electroencephalographic signals, coupled with their low signal-to-noise ratio, lead to variations in amplitude, time period, and latency in the patterns of event-related potentials across different trials, sessions, days and subjects. Conventional feature extraction and machine learning algorithms are not designed to handle these differences, requiring the development of methods that can address these variations. Here, we propose a novel lightweight deep neural network for event-related potential classification, consisting of three modules. In this model, we have a spatio-temporal module that learns local features simultaneously across channels and time points. Following this, there's a component extractor module comprising depthwise convolutions, inspired by mixed depthwise convolutions, to capture the event-related potential characteristics with different temporal durations. Lastly, an advanced temporal layer addresses event-related potential shape and scale variations using deformable convolutions. We conducted experiments on event-related potential detection in a subject-independent scenario using one error-related negativity potential dataset and three perturbation-evoked potential datasets. Comparisons were made with established methods including two conventional machine learning algorithms and three well-known deep learning architectures, demonstrating that our model outperformed them in terms of classification accuracy and parameter efficiency. In our analysis, we aimed to understand the model's performance using gradient-weighted class activation mapping and t-distributed stochastic neighbor embedding. These methods facilitated the visualization and interpretation of our model's effectiveness, providing insights into its relationship with the neuroscientific characteristics of event-related potentials.

Category-Level Pose Estimation and Iterative Refinement for Monocular RGB-D Image

Category-level pose estimation is proposed to predict the 6D pose of objects under a specific category and has wide applications in fields such as robotics, virtual reality, and autonomous driving. With the development of VR/AR technology, pose estimation has gradually become a research hotspot in 3D scene understanding. However, most methods fail to fully utilize geometric and color information to solve intra-class shape variations, which leads to inaccurate prediction results. To solve the above problems, we propose a novel pose estimation and iterative refinement network, use an attention mechanism to fuse multi-modal information to obtain color features after a coordinate transformation, and design iterative modules to ensure the accuracy of object geometric features. Specifically, we use an encoder-decoder architecture to implicitly generate a coarse-grained initial pose and refine it through an iterative refinement module. In addition, due to the differences between rotation and position estimation, we design a multi-head pose decoder that utilizes the local geometry and global features. Finally, we design a transformer-based coordinate transformation attention module to extract pose-sensitive features from RGB images and supervise color information by correlating point cloud features in different coordinate systems. We train and test our network on the synthetic dataset CAMERA25 and the real dataset REAL275. Experimental results show that our method achieves state-of-the-art performance on multiple evaluation metrics.