Multi-scale Feature Aggregation Research Articles

A Lightweight Face Detector via Bi-Stream Convolutional Neural Network and Vision Transformer

Lightweight convolutional neural networks are widely used for face detection due to their ability to learn local representations through spatial induction bias and translational invariance. However, convolutional face detectors have limitations in detecting faces under challenging conditions like occlusion, blurring, or changes in facial poses, primarily attributed to fixed-size receptive fields and a lack of global modeling. Transformer-based models have advantages on learning global representations but are insensitive to capture local patterns. To address these limitations, we propose an efficient face detector that combines convolutional neural network and transformer architectures. We introduce a bi-stream structure that integrates convolutional neural network and transformer blocks within the backbone network, enabling the preservation of local pattern features and the extraction of global context. To further preserve the local details captured by convolutional neural networks, we propose a feature enhancement convolution block in a hierarchical backbone structure. Additionally, we devise a multiscale feature aggregation module to enhance obscured and blurred facial features. Experimental results demonstrate that our method has achieved improved lightweight face detection accuracy with an average precision of 95.30%, 94.20%, and 87.56% across the easy, medium, and hard subdatasets of WIDER FACE, respectively. Therefore, we believe our method will be a useful supplement to the collection of current artificial intelligence models and benefit the engineering applications of face detection.

MSFA-Net: A Multiscale Feature Aggregation Network for Semantic Segmentation of Historical Building Point Clouds

In recent years, research on the preservation of historical architecture has gained significant attention, where the effectiveness of semantic segmentation is particularly crucial for subsequent repair, protection, and 3D reconstruction. Given the sparse and uneven nature of large-scale historical building point cloud scenes, most semantic segmentation methods opt to sample representative subsets of points, often leading to the loss of key features and insufficient segmentation accuracy of architectural components. Moreover, the geometric feature information at the junctions of components is cluttered and dense, resulting in poor edge segmentation. Based on this, this paper proposes a unique semantic segmentation network design called MSFA-Net. To obtain multiscale features and suppress irrelevant information, a double attention aggregation module is first introduced. Then, to enhance the model’s robustness and generalization capabilities, a contextual feature enhancement and edge interactive classifier module are proposed to train edge features and fuse the context data. Finally, to evaluate the performance of the proposed model, experiments were conducted on a self-curated ancient building dataset and the S3DIS dataset, achieving OA values of 95.2% and 88.7%, as well as mIoU values of 86.2% and 71.6%, respectively, further confirming the effectiveness and superiority of the proposed method.

Multi-scale feature aggregation and fusion network with self-supervised multi-level perceptual loss for textures preserving low-dose CT denoising

Objective. The textures and detailed structures in computed tomography (CT) images are highly desirable for clinical diagnosis. This study aims to expand the current body of work on textures and details preserving convolutional neural networks for low-dose CT (LDCT) image denoising task. Approach. This study proposed a novel multi-scale feature aggregation and fusion network (MFAF-net) for LDCT image denoising. Specifically, we proposed a multi-scale residual feature aggregation module to characterize multi-scale structural information in CT images, which captures regional-specific inter-scale variations using learned weights. We further proposed a cross-level feature fusion module to integrate cross-level features, which adaptively weights the contributions of features from encoder to decoder by using a spatial pyramid attention mechanism. Moreover, we proposed a self-supervised multi-level perceptual loss module to generate multi-level auxiliary perceptual supervision for recovery of salient textures and structures of tissues and lesions in CT images, which takes advantage of abundant semantic information at various levels. We introduced parameters for the perceptual loss to adaptively weight the contributions of auxiliary features of different levels and we also introduced an automatic parameter tuning strategy for these parameters. Main results. Extensive experimental studies were performed to validate the effectiveness of the proposed method. Experimental results demonstrate that the proposed method can achieve better performance on both fine textures preservation and noise suppression for CT image denoising task compared with other competitive convolutional neural network (CNN) based methods. Significance. The proposed MFAF-net takes advantage of multi-scale receptive fields, cross-level features integration and self-supervised multi-level perceptual loss, enabling more effective recovering of fine textures and detailed structures of tissues and lesions in CT images.

DFA-Net: Dual multi-scale feature aggregation network for vessel segmentation in X-ray digital subtraction angiography

Even though deep learning is fascinated in fields of coronary vessel segmentation in X-ray angiography and achieves prominent progresses, most of those models probably bring high false and missed detections due to indistinct contrast between coronary vessels and background, especially for tiny sub-branches. Image improvement technique is able to better such contrast, while boosting extraneous information, e.g., other tissues with similar intensities and noise. If incorporating features derived from original and enhanced images, the segmentation performance is improved because those images comprise complementary information from different contrasts. Accordingly, inspired from advantages of contrast improvement and encoding-decoding architecture, a dual multi-scale feature aggregation network (named DFA-Net) is introduced for coronary vessel segmentation in digital subtraction angiography (DSA). DFA-Net integrates the contrast improvement using exponent transformation into a semantic segmentation network that individually accepts original and enhanced images as inputs. Through parameter sharing, multi-scale complementary features are aggregated from different contrasts, which strengthens leaning capabilities of networks, and thus achieves an efficient segmentation. Meanwhile, a risk cross-entropy loss is enforced on the segmentation, for availably decreasing false negatives, which is incorporated with Dice loss for joint optimization of the proposed strategy during training. Experimental results demonstrate that DFA-Net can not only work more robustly and effectively for DSA images under diverse conditions, but also achieve better performance, in comparison with state-of-the-art methods. Consequently, DFA-Net has high fidelity and structure similarity to the reference, providing a way for early diagnosis of cardiovascular diseases.

Prediction of recurrence-free survival in lung adenocarcinoma based on self-supervised pre-training and multi-task learning

Computed tomography (CT) imaging is a vital tool for the diagnosis and assessment of lung adenocarcinoma, and using CT images to predict the recurrence-free survival (RFS) of lung adenocarcinoma patients post-surgery is of paramount importance in tailoring postoperative treatment plans. Addressing the challenging task of accurate RFS prediction using CT images, this paper introduces an innovative approach based on self-supervised pre-training and multi-task learning. We employed a self-supervised learning strategy known as "image transformation to image restoration" to pretrain a 3D-UNet network on publicly available lung CT datasets to extract generic visual features from lung images. Subsequently, we enhanced the network's feature extraction capability through multi-task learning involving segmentation and classification tasks, guiding the network to extract image features relevant to RFS. Additionally, we designed a multi-scale feature aggregation module to comprehensively amalgamate multi-scale image features, and ultimately predicted the RFS risk score for lung adenocarcinoma with the aid of a feed-forward neural network. The predictive performance of the proposed method was assessed by ten-fold cross-validation. The results showed that the consistency index (C-index) of the proposed method for predicting RFS and the area under curve (AUC) for predicting whether recurrence occurs within three years reached 0.691 ± 0.076 and 0.707 ± 0.082, respectively, and the predictive performance was superior to that of existing methods. This study confirms that the proposed method has the potential of RFS prediction in lung adenocarcinoma patients, which is expected to provide a reliable basis for the development of individualized treatment plans.

Meta-learning based infrared ship object detection model for generalization to unknown domains

Infrared images exhibit considerable variations in probability distributions, stemming from the utilization of distinct infrared sensors and the influence of diverse environmental conditions. The variations pose great challenges for deep learning models to detect ship objects and adapt to unseen maritime environments. To address the domain shift problem, we propose an end-to-end infrared ship object detection model based on meta-learning neural network to improve domain adaptation for target domain where data is not available at training phase. Different from existing domain generalization methods, the novelty of our model lies in the effective exploitation of meta-learning and domain adaptation, ensuring that the extracted domain-independent features are meaningful and domain-invariant at the semantic level. Firstly, a double gradient-based meta-learning algorithm is designed to solve the common optimal descent direction between different domains through two gradient updates in the inner and outer loops. The algorithm enables extraction of domain-invariant features from the pseudo-source and pseudo-target domain data. Secondly, a domain discriminator with dynamic-weighted gradient reversal layer (DWGRL) is designed to accurately classify domain-invariant features and provide additional global supervision information. Finally, a multi-scale feature aggregation method is proposed to improve the extraction of multi-scale domain-invariant features. It can effectively fuse local features at different scales and global features of targets. Extensive experimental results conducted in real nighttime water surface scenes demonstrate that the proposed model achieves very high detection accuracy on target domain data, even no target domain data was used during the training phase. Compared to the existing methods, our method not only improves the detection accuracy of infrared ships by 18%, but also exhibits the smallest standard deviation with a value of 0.93, indicating its superior generalization performance.

Transformer-CNN hybrid network for crowd counting

Efficient feature representation is the key to improving crowd counting performance. CNN and Transformer are the two commonly used feature extraction frameworks in the field of crowd counting. CNN excels at hierarchically extracting local features to obtain a multi-scale feature representation of the image, but it struggles with capturing global features. Transformer, on the other hand, could capture global feature representation by utilizing cascaded self-attention to capture remote dependency relationships, but it often overlooks local detail information. Therefore, relying solely on CNN or Transformer for crowd counting has certain limitations. In this paper, we propose the TCHNet crowd counting model by combining the CNN and Transformer frameworks. The model employs the CMT (CNNs Meet Vision Transformers) backbone network as the Feature Extraction Module (FEM) to hierarchically extract local and global features of the crowd using a combination of convolution and self-attention mechanisms. To obtain more comprehensive spatial local information, an improved Progressive Multi-scale Learning Process (PMLP) is introduced into the FEM, guiding the network to learn at different granularity levels. The features from these three different granularity levels are then fed into the Multi-scale Feature Aggregation Module (MFAM) for fusion. Finally, a Multi-Scale Regression Module (MSRM) is designed to handle the multi-scale fused features, resulting in crowd features rich in high-level semantics and low-level detail. Experimental results on five benchmark datasets demonstrate that TCHNet achieves highly competitive performance compared to some popular crowd counting methods.

OmiQnet: Multiscale feature aggregation convolutional neural network for omnidirectional image assessment

OmiQnet: Multiscale feature aggregation convolutional neural network for omnidirectional image assessment

CMG-Net: Robust Normal Estimation for Point Clouds via Chamfer Normal Distance and Multi-Scale Geometry

This work presents an accurate and robust method for estimating normals from point clouds. In contrast to predecessor approaches that minimize the deviations between the annotated and the predicted normals directly, leading to direction inconsistency, we first propose a new metric termed Chamfer Normal Distance to address this issue. This not only mitigates the challenge but also facilitates network training and substantially enhances the network robustness against noise. Subsequently, we devise an innovative architecture that encompasses Multi-scale Local Feature Aggregation and Hierarchical Geometric Information Fusion. This design empowers the network to capture intricate geometric details more effectively and alleviate the ambiguity in scale selection. Extensive experiments demonstrate that our method achieves the state-of-the-art performance on both synthetic and real-world datasets, particularly in scenarios contaminated by noise. Our implementation is available at https://github.com/YingruiWoo/CMG-Net_Pytorch.

YOLO-DCNet

Enhanced processors empower edge devices like smartphones for human detection, yet their application is constrained by algorithmic efficiency and precision. This paper introduces YOLO-DCNet, a lightweight neural network detector built upon YOLOv7-tiny. Incorporating a dynamic multi-head structural re-parameterization (DMSR) module within its backbone network enables effective processing of the features utilized in the model. To improve multi-scale feature aggregation, the model integrates a channel information compression and linear mapping (CLM) module into its feature pyramid architecture. Moreover, the optimization of training and inference performance is achieved by employing RepVGG blocks between the main computational modules of the model. Experimental data reveal that the enhanced YOLOv7-tiny model achieves a 31.7% faster inference speed and marginal gains of 0.7% in mAP@0.5 and 0.5% in mAP@0.5:0.95 over the original. This underscores the model's improved performance and applicability for real-time human detection on edge devices across diverse applications.

Enhancing identification for person search with multi-scale multi-grained representation learning

Person Search aims to simultaneously address Person Detection and Person Re-ID. There are various challenges in person search such as significant scale variations, occlusions, and partial instances. In this paper, we propose a Multi-Scale Multi-Grained (MSMG) sequential network for end-to-end person search, intended to alleviate these issues. To generate re-id representations robust to scale changes, MSMG leverages multi-scale RoI features and aggregates them with a proposed Multi-Scale feature Aggregation Encoder (MSAE). In this way, the aggregated multi-scale re-id features are enriched with more semantic information and detailed information, thereby being more discriminative for identification. Moreover, to produce re-id representations more robust to occlusions and partial instances, MSMG introduces a Multi-Grained feature Learning Decoder (MGLD) focused on multi-grained feature learning. MGLD adaptively decodes multi-grained re-id representations with more accurate semantic information through a regional deformable cross-attention module. Finally, the multi-scale multi-grained re-id representation substantially improves the identification accuracy under challenging cases. Through comprehensive experiments, we demonstrate that our method achieves state-of-the-art performance on two benchmark datasets. On the challenging PRW benchmark, MSMG obtains the best-reported results with a mean average precision (mAP) score of 61.3%.

Exhaust Gas Temperature Prediction of Aero-Engine via Enhanced Scale-Aware Efficient Transformer

This research introduces the Enhanced Scale-Aware efficient Transformer (ESAE-Transformer), a novel and advanced model dedicated to predicting Exhaust Gas Temperature (EGT). The ESAE-Transformer merges the Multi-Head ProbSparse Attention mechanism with the established Transformer architecture, significantly optimizing computational efficiency and effectively discerning key temporal patterns. The incorporation of the Multi-Scale Feature Aggregation Module (MSFAM) further refines 2 s input and output timeframe. A detailed investigation into the feature dimensionality was undertaken, leading to an optimized configuration of the model, thereby improving its overall performance. The efficacy of the ESAE-Transformer was rigorously evaluated through an exhaustive ablation study, focusing on the contribution of each constituent module. The findings showcase a mean absolute prediction error of 3.47∘R, demonstrating strong alignment with real-world environmental scenarios and confirming the model’s accuracy and relevance. The ESAE-Transformer not only excels in predictive accuracy but also sheds light on the underlying physical processes, thus enhancing its practical application in real-world settings. The model stands out as a robust tool for critical parameter prediction in aero-engine systems, paving the way for future advancements in engine prognostics and diagnostics.

Speaker-turn aware diarization for speech-based cognitive assessments.

Speaker diarization is an essential preprocessing step for diagnosing cognitive impairments from speech-based Montreal cognitive assessments (MoCA). This paper proposes three enhancements to the conventional speaker diarization methods for such assessments. The enhancements tackle the challenges of diarizing MoCA recordings on two fronts. First, multi-scale channel interdependence speaker embedding is used as the front-end speaker representation for overcoming the acoustic mismatch caused by far-field microphones. Specifically, a squeeze-and-excitation (SE) unit and channel-dependent attention are added to Res2Net blocks for multi-scale feature aggregation. Second, a sequence comparison approach with a holistic view of the whole conversation is applied to measure the similarity of short speech segments in the conversation, which results in a speaker-turn aware scoring matrix for the subsequent clustering step. Third, to further enhance the diarization performance, we propose incorporating a pairwise similarity measure so that the speaker-turn aware scoring matrix contains both local and global information across the segments. Evaluations on an interactive MoCA dataset show that the proposed enhancements lead to a diarization system that outperforms the conventional x-vector/PLDA systems under language-, age-, and microphone-mismatch scenarios. The results also show that the proposed enhancements can help hypothesize the speaker-turn timestamps, making the diarization method amendable to datasets without timestamp information.

Quantitative characterization of fractures and holes in core rolling scan images based on the MFAPNet deep learning model

The exploration and comprehensive assessment of fractured-vuggy reservoir information have perennially constituted focal points and challenges within the domain of oil and gas reservoir evaluation. The verification of geological phenomena, identification of various fracture and hole types, and the quantitative characterization thereof currently present pressing challenges. This study meticulously examines the deep carbonate reservoirs within the Dengying Formation in the Penglai gas region of the Sichuan Basin. The Core Rolling Scan images reveal five discernible features: unfilled holes, filled holes, filled fractures, open fractures, and algae. The analysis pinpoints three primary challenges in semantic segmentation recognition: the amalgamation of feature scales, class imbalance, and the scarcity of datasets with substantial sample sizes. To address these challenges, this paper introduces a Multi-Scale Feature Aggregation Pyramid Network model (MFAPNet), achieving a pixel accuracy of 68.04% in recognizing the aforementioned five types. Lastly, the model is employed in calculating core porosity, exposing a scaling relationship between wellbore image porosity and core porosity ranging from 1.5 to 3 times. To a certain extent, it reveals the correlation between the wellbore image logging data and the actual formation of the Dengying Formation in the Penglai Gas Field of the Sichuan Basin, and also provides a basis for the subsequent logging evaluation of the formation. The partial code and CHA355 dataset are publicly available at https://github.com/zyng886/MFAPNet.

Learning to Restore Compressed Point Cloud Attribute: A Fully Data-Driven Approach and A Rules-Unrolling-Based Optimization.

The emergence of holographic media drives the standardization of Geometry-based Point Cloud Compression (G-PCC) to sustain networked service provisioning. However, G-PCC inevitably introduces visually annoying artifacts, degrading the quality of experience (QoE). This work focuses on restoring G-PCC compressed point cloud attributes, e.g., RGB colors, to which fully data-driven and rules-unrolling-based post-processing filters are studied. At first, as compressed attributes exhibit nested blockiness, we develop a learning-based sample adaptive offset (NeuralSAO), which leverages a neural model using multiscale feature aggregation and embedding to characterize local correlations for quantization error compensation. Later, given statistically Gaussian distributed quantization noise, we suggest the utilization of a bilateral filter with Gaussian kernels to weigh neighbors by jointly considering their geometric and photometric contributions for restoration. Since local signals often present varying distributions, we propose estimating the smoothing parameters of the bilateral filter using an ultra-lightweight neural model. Such a bilateral filter with learnable parameters is called NeuralBF. The proposed NeuralSAO demonstrates the state-of-art restoration quality improvement, e.g., >20% BD-BR (Bjøntegaard delta rate) reduction over G-PCC on solid points clouds. However, NeuralSAO is computationally intensive and may suffer from poor generalization. On the other hand, although NeuralBF only achieves half of the gains of NeuralSAO, it is lightweight and exhibits impressive generalization across various samples. This comparative study between the data-driven large-scale NeuralSAO and the rules-unrolling-based small-scale NeuralBF helps to understand the capacity (i.e., performance, complexity, generalization) of underlying filters in terms of the quality restoration for compressed point cloud attribute.

MM-Net: A MixFormer-Based Multi-Scale Network for Anatomical and Functional Image Fusion.

Anatomical and functional image fusion is an important technique in a variety of medical and biological applications. Recently, deep learning (DL)-based methods have become a mainstream direction in the field of multi-modal image fusion. However, existing DL-based fusion approaches have difficulty in effectively capturing local features and global contextual information simultaneously. In addition, the scale diversity of features, which is a crucial issue in image fusion, often lacks adequate attention in most existing works. In this paper, to address the above problems, we propose a MixFormer-based multi-scale network, termed as MM-Net, for anatomical and functional image fusion. In our method, an improved MixFormer-based backbone is introduced to sufficiently extract both local features and global contextual information at multiple scales from the source images. The features from different source images are fused at multiple scales based on a multi-source spatial attention-based cross-modality feature fusion (CMFF) module. The scale diversity of the fused features is further enriched by a series of multi-scale feature interaction (MSFI) modules and feature aggregation upsample (FAU) modules. Moreover, a loss function consisting of both spatial domain and frequency domain components is devised to train the proposed fusion model. Experimental results demonstrate that our method outperforms several state-of-the-art fusion methods on both qualitative and quantitative comparisons, and the proposed fusion model exhibits good generalization capability. The source code of our fusion method will be available at https://github.com/yuliu316316.

Hider: A Hyperspectral Image Denoising Transformer With Spatial-Spectral Constraints for Hybrid Noise Removal.

Hyperspectral image (HSI) qualities are limited by a mixture of Gaussian noise, impulse noise, stripes, and deadlines during the sensor imaging process, resulting in weak application performance. To enhance HSI qualities, methods based on convolutional neural networks have been successively applied to restore clean data from the observed data. However, the architecture of these methods lacks spectral and spatial constraints, and the convolution operators have limited receptive fields and inflexible model inferences. Thus, in this study, we propose an efficient end-to-end transformer, named HSI denoising transformer (Hider), for mixed HSI noise removal. First, a U-shaped 3-D transformer architecture is built for multiscale feature aggregation. Second, a multihead global spectral attention module within the spectral transformer block is designed to excavate information in different spectral patterns. Finally, an additional locally enhanced cross-spatial attention module within the spatial-spectral transformer block is constructed to build the long-range spatial relationship to avoid the high computational complexity of global spatial self-attention. Through the imposition of global correlations along spectrum and spatial self-similarity constraints on the transformer, our proposed Hider aims to capture long-range spatial contextual information and cluster objects with the same spectral pattern for HSI denoising. To verify the effectiveness and efficiency of Hider, we conducted extensive simulated and real experiments. The denoising results on both simulated and real-world datasets show that Hider achieves superior evaluation metrics and visual assessments compared with other state-of-the-art methods.

Rolling bearing fault diagnosis method based on MTF-MFACNN

A rolling bearing fault diagnosis method based on the Markov transition field (MTF) and multi-scale feature aggregation convolutional neural network (MFACNN) is proposed to address the problems of excessive parameter number, slow training speed, and insufficient generalization of traditional CNNs. Firstly, the original vibration signal is input into the MTF and converted into two-dimensional images with time correlation. Then, in order to effectively aggregate feature information at different scales and levels, a MFA module is presented to capture rich information from feature maps at different scales and assign different weights to these features for fusion. Secondly, while ensuring the lightweight of the model, utilizing feature information of different resolutions, a lightweight feature fusion module is put forward to fuse multiple feature maps together to improve the performance and efficiency of the model. On this basis, an MFACNN model is constructed. Finally, the two-dimensional images are input into MTF-MFACNN and experimentally validated using two different datasets. The results show that the proposed method has faster calculation speed, higher fault recognition accuracy, and stronger generalization performance compared to other methods.

MAEU‐NET: A novel supervised architecture for brain tumor segmentation

AbstractA brain tumor is an abnormal growth of cells that damages the neural system and may lead to severe conditions. These cells are irregular in shape and size, and as a result, the manual analysis of these cells from magnetic resonance (MR) images is a cumbersome and time‐consuming task. To alleviate these issues, a novel U‐Net inspired MAEU‐Net (Multiscale context Aggregation and Edge activation U‐Net) is implemented to detect the brain tumor pixels from MR images more precisely. The encoder block of the proposed model is improved by integrating with multiscale feature aggregation and convolution block attention modules (CBAM) enabling the model to extract both pertinent low‐level and high‐level features. Further, the extracted features are subjected to a parallel pooling module to elevate the features that extract the tumor pixels. In addition to this, the output features of the last encoder are fed to a parallel dilated context aggregation module to mitigate the hyper‐parameter size and further elevate the receptive field strength of the proposed model. Apart from this, edge‐enhanced modules are incorporated with skip connections which concatenate the feature maps of the decoder with respective edge‐enhanced encoder feature maps. As a result, this module increases the aspect ratio and reduces blurry edge problems. To illustrate the performance of the proposed MAEU‐Net, it is trained and tested using Kaggle LGG, BraTS 2018, and BraTS 2021 publicly available datasets. The proposed model outperformed the previously existing models by achieving the test loss, dice coefficient, sensitivity, and specificity of 0.0590, 0.901, 0.942, and 0.998, respectively. Simultaneously, the proposed model is quantitatively analyzed by performing ablation and comparative studies and presented that the proposed MAEU‐Net is superior to other previous existing models like U‐Net++, SegNet, etc.

MADANet: A Lightweight Hyperspectral Image Classification Network with Multiscale Feature Aggregation and a Dual Attention Mechanism

Hyperspectral remote sensing images, with their continuous, narrow, and rich spectra, hold distinct significance in the precise classification of land cover. Deep convolutional neural networks (CNNs) and their variants are increasingly utilized for hyperspectral classification, but solving the conflict between the number of model parameters, performance, and accuracy has become a pressing challenge. To alleviate this problem, we propose MADANet, a lightweight hyperspectral image classification network that combines multiscale feature aggregation and a dual attention mechanism. By employing depthwise separable convolution, multiscale features can be extracted and aggregated to capture local contextual information effectively. Simultaneously, the dual attention mechanism harnesses both channel and spatial dimensions to acquire comprehensive global semantic information. Ultimately, techniques such as global average pooling (GAP) and full connection (FC) are employed to integrate local contextual information with global semantic knowledge, thereby enabling the accurate classification of hyperspectral pixels. The results from the experiments conducted on representative hyperspectral images demonstrate that MADANet not only attains the highest classification accuracy but also maintains significantly fewer parameters compared to the other methods. Experimental results show that our proposed framework significantly reduces the number of model parameters while still achieving the highest classification accuracy. As an example, the model has only 0.16 M model parameters in the Indian Pines (IP) dataset, but the overall accuracy is as high as 98.34%. Similarly, the framework achieves an overall accuracy of 99.13%, 99.17%, and 99.08% on the University of Pavia (PU), Salinas (SA), and WHU Hi LongKou (LongKou) datasets, respectively. This result exceeds the classification accuracy of existing state-of-the-art frameworks under the same conditions.