Method and device for detecting the number of layers of two-dimensional materials under a microscope based on deep learning

Abstract

The present invention discloses a method and device for detecting the number of layers of two-dimensional materials under a microscope based on deep learning. The method captures high-resolution RGB images of deposited two-dimensional materials through an optical microscope and performs preprocessing. After transfer learning based on the pre-trained Mask2Former model, it is used for the training of two-dimensional material images, and the exponential moving average (EMA) technique is used to smooth the parameters of the model. The trained Mask2Former model is applied to newly acquired two-dimensional material images for segmentation to generate segmentation masks for each region, and the number of layers of each region is predicted based on the masks. The area ratio of each layer region is calculated according to the pixel point ratio to obtain the percentage parameter of each layer. The present invention is applicable to a variety of 2DP materials and can be widely applied to the research and application of 2DP materials in the fields of flexible electronics, biosensing, drug release, optoelectronic devices, etc.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, in particular to a method and device for detecting the number of layers of two-dimensional materials under a microscope based on deep learning. Background Art

[0002] Two-dimensional materials have received extensive attention in the field of materials science due to their unique properties. Among them, two-dimensional polymers (2DPs), as emerging members of the two-dimensional material family, exhibit excellent application prospects in fields such as flexible electronics, biosensing, and drug release. The number of layers of 2DPs has an important impact on their electronic properties, optical properties, etc., especially in applications in fields such as semiconductors and optoelectronic devices. Therefore, accurately and quickly identifying the number of layers of two-dimensional polymers is of great significance in materials science and industrial production, and accurately detecting its number of layers has become a key issue in research. Traditional detection methods, such as transmission electron microscopy (TEM) and atomic force microscopy (AFM), although able to provide accurate layer number information, are complex to operate and costly. In recent years, the application of deep learning technology in image analysis has made remarkable progress, especially in the field of computer vision, and has gradually become an effective solution. With the development of deep learning, models such as convolutional neural networks (CNNs) have made remarkable progress in the field of image recognition, but still face three major challenges when processing complex 2DP optical microscopy images:

[0003] Complexity of the unique optical interference pattern of 2DP: 2DP materials exhibit unique interference fringes and color changes under an optical microscope. These features are related to the number of layers of the material, but traditional CNN models are difficult to effectively capture such complex optical characteristics;

[0004] Scarcity of labeled data for 2DP samples: The acquisition cost of high-quality 2DP layer number labeled data is high and the quantity is limited, making it difficult to support the effective training of complex deep learning models;

[0005] Difficulty in integrating multi-scale features of 2DP samples: 2DP materials often exhibit gradient changes in the transition regions between different numbers of layers, and it is necessary to consider both local microstructures and global context information simultaneously, which poses higher requirements for model design. For example, the accuracy is insufficient when processing complex and blurred images, and there is a lack of a certain degree of robustness. Therefore, it is necessary to develop a new method that combines the latest achievements of deep learning to improve the detection accuracy and stability.

[0006] Currently, there is no deep learning method specifically optimized for the characteristics of 2DP materials that can effectively overcome the above three key technical problems at the same time, and there is an urgent need to develop a dedicated solution. Summary of the Invention

[0007] To address the deficiencies in the prior art, the present invention provides a method and apparatus for detecting the number of layers of two-dimensional materials under a microscope based on deep learning. In particular, it is for detecting the number of layers of two-dimensional polymers (2DP) under a microscope based on the MaskEMA-2DP model. This method innovatively designs an enhanced Mask model specifically for the characteristics of 2DP. Based on Mask2Former in the Transformer architecture, it innovatively integrates the application of transfer learning and exponential moving average (EMA) algorithm specifically for the material characteristics of 2DP, overcoming the above three major technical problems. It can accurately detect the number of layers of 2DP through optical microscope images and has high accuracy, robustness, and efficiency.

[0008] The technical solution adopted by the present invention to solve the above technical problems is: a method for detecting the number of layers of two-dimensional materials under a microscope based on deep learning, the method comprising the following steps:

[0009] a) Sample preparation: Mix the two-dimensional material with a solvent. Under the action of the solvent, the two-dimensional material expands and is layered under the action of shear force to obtain several layers of two-dimensional material, and deposit the two-dimensional material on a substrate;

[0010] b) Image acquisition: Take a high-resolution RGB image of the deposited two-dimensional material through an optical microscope, and use it for training the deep learning model after preprocessing;

[0011] c) Model training: Based on the pre-trained Mask2Former model, perform transfer learning and then use it for training the two-dimensional material images, and use the exponential moving average EMA technique to smooth the parameters of the model;

[0012] d) Layer number detection: Apply the trained Mask2Former model to the newly acquired two-dimensional material image for segmentation, generate a segmentation mask for each region, and predict the number of layers of each region according to the mask;

[0013] e) Layer quantification analysis: Calculate the area ratio of each layer number region according to the pixel point ratio to obtain the percentage parameter of each layer number.

[0014] Further, the image preprocessing includes image denoising, multi-scale contrast enhancement, brightness adaptive correction, data normalization, and data annotation.

[0015] Further, the denoising is a denoising method that combines non-local mean filtering and wavelet transform to suppress the interference patterns in the two-dimensional material image while retaining the key information related to the number of layers; the multi-scale contrast enhancement uses multi-scale adaptive histogram equalization technology to locally optimize the feature differences in different layer regions of the two-dimensional material, making the layer boundaries clearer; the brightness adaptive correction is based on the optical characteristics of the two-dimensional material and uses an adaptive brightness correction algorithm; the data annotation uses SAM to assist in annotating the two-dimensional material image to obtain the segmentation results of different layer regions.

[0016] Further, the Mask2Former framework is based on the Transformer architecture and performs image segmentation through multiple layers of decoders to process the global and local features in the two-dimensional material image.

[0017] Further, the method for detecting the number of layers is used for detecting the number of layers of two-dimensional polymer 2DP under a microscope.

[0018] Further, the layer number prediction results of the Mask2Former model are displayed in the image in a color-coded manner, and different layer regions are marked with different colors.

[0019] Further, the layer quantification analysis outputs the percentage of each layer region and displays the distribution of different layers through statistical charts.

[0020] Further, during the model training process, for the key features of two-dimensional material layer detection, the following feature weighting mechanism is adopted:

[0021] 1) Assign a higher EMA weight to the RGB information feature channels of the number of layers of two-dimensional materials;

[0022] 2) Dynamically optimize and adjust the weights of the RGB information feature channels according to the training results;

[0023] 3) Assign a higher EMA weight to the high-certainty regions confirmed by the Mask2Former model.

[0024] In a second aspect, the present invention also provides a device for detecting the number of layers of two-dimensional materials under a microscope based on deep learning, including a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it implements the method for detecting the number of layers of two-dimensional materials under a microscope based on deep learning as described above.

[0025] In a third aspect, the present invention also provides a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, it implements the method for detecting the number of layers of two-dimensional materials under a microscope based on deep learning as described above.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] Improve detection accuracy: The MaskEMA-2DP model of the present invention is specifically optimized for the 2DP optical characteristics, significantly improving the accuracy of layer detection. The self-attention mechanism of Mask2Former can deeply capture the global information in the image, thereby accurately segmenting and identifying different hierarchical regions, overcoming the limitations of traditional methods in complex images.

[0028] Reduce training time: By introducing transfer learning, the present invention effectively reduces the dependence on large-scale labeled datasets and accelerates the convergence speed of the model. In addition, the EMA technology makes the model more stable during training, thereby reducing the training time and ensuring high-precision detection results can be obtained in a short time. Through the three-stage transfer learning strategy and the 2DP feature-weighted EMA technology, the present invention significantly reduces the demand for labeled data. Experiments have shown that only 100-200 labeled 2DP images are required to achieve good performance, reducing the data demand by more than 70% compared with traditional methods.

[0029] Enhance robustness and stability: The EMA technology can smooth the parameter changes during training, enabling the model to maintain stable performance when facing different brightness, contrast changes, or other experimental condition changes. This means that the method has strong adaptability in practical applications and can cope with changing experimental environments.

[0030] The MaskEMA-2DP model of the present invention has strong adaptability to the changes of 2DP samples under different lighting conditions. Experiments have verified that under the conditions of brightness change of ±10% and contrast change of ±10%, the model performance degradation does not exceed 3%, far superior to the existing methods.

[0031] Reduce manual intervention: Traditional layer detection methods usually rely on manual annotation, which is labor-intensive and error-prone. By using the deep learning model of the present invention, image segmentation and layer detection can be automatically completed, greatly reducing manual intervention and annotation time and improving work efficiency. Brief Description of the Drawings

[0032] Figure 1 It is a schematic diagram of the sample preparation of the present invention.

[0033] Figure 2 It is a schematic diagram of the model and training framework of the present invention.

[0034] Figure 3 It is a schematic diagram of the image segmentation and layer detection results of the present invention.

[0035] Figure 4Schematic diagram of the results of improving the present invention model by adding EMA and transfer learning.

[0036] Figure 5 Schematic diagram showing the robustness of the model after changing the brightness and contrast of the present invention.

[0037] Figure 6 Structural diagram of a device for detecting the number of layers of two-dimensional materials under a microscope based on deep learning provided by the present invention. Detailed implementation manners

[0038] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described below with reference to the accompanying drawings and implementation cases. It should be understood that the specific implementation cases described herein are only used to explain the present invention and are not used to limit the present invention.

[0039] The present invention discloses a method for detecting the number of layers of two-dimensional materials under a microscope based on deep learning. The specific implementation process of this method is as follows:

[0040] 1. Sample preparation

[0041] As shown in Figure 1 the left figure in: The figure illustrates the experimental process of using a high-shear mixing device for the exfoliation of two-dimensional materials. The high-shear stirrer gradually exfoliates the layered crystals of the raw materials into single-layer or few-layer two-dimensional materials through strong stirring. The solution exfoliation process of 2DP includes mixing the 2DP material with a solvent (such as trifluoroacetic acid TFA). TFA causes the 2DP material to expand and delaminate under the action of shear force, generating the required single-layer or few-layer materials. The device in the figure includes a stirrer 1 and a liquid container 2.

[0042] As shown in Figure 1 the middle figure in: In the experiment, a pipette is used to evenly distribute the stirred solution onto a pre-cleaned and treated SiO2 / Si substrate, which helps with subsequent image analysis.

[0043] As shown in Figure 1 the right figure in: A microscope is used to observe the deposited 2DP sample and capture high-resolution images, providing data for model training and layer analysis. The detailed structural information of the two-dimensional material is captured through high-resolution RGB images, providing a basis for subsequent segmentation and layer detection.

[0044] 2. Image acquisition and preprocessing

[0045] Microscope image acquisition: Use an optical microscope to capture high-resolution images of two-dimensional material samples, with a resolution set to 1000 dpi and a size of 512x512 pixels. The microscope needs to ensure that the images can clearly present the multi-layer structure of the material and have enough details for precise analysis.

[0046] Preprocessing steps:

[0047] 2DP-specific denoising: A denoising method that combines non-local means filtering and wavelet transform, specifically suppressing common interference patterns in 2DP images while retaining key information related to the number of layers;

[0048] Multi-scale contrast enhancement: Using multi-scale adaptive histogram equalization technology, locally optimizing the characteristics of different layer regions in 2DP to make the layer boundaries clearer;

[0049] Adaptive brightness correction: Adjust the brightness according to the image quality to adapt to images under different lighting conditions, enabling the model to accurately identify the number of layers in various environments. An adaptive brightness correction algorithm designed based on the optical characteristics of 2DP materials can automatically adjust the optimal viewing brightness according to material characteristics, improving the distinguishability of different layer regions;

[0050] Data normalization: Normalize the image to unify the pixel value range of the image, so that the deep learning model can better learn and adapt to image features.

[0051] Data annotation: Use the Segment Anything Model (SAM) to assist in annotating two-dimensional polymer images to obtain the segmentation results of regions with different numbers of layers, where the regions include single-layer, double-layer, triple-layer and more layers;

[0052] 3. Model framework and training

[0053] As Figure 2 shown, the MaskEMA-2DP model contains multiple innovative components specifically designed for 2DP materials. The model includes several main parts:

[0054] Backbone: First, the image passes through the backbone for feature extraction to generate a low-resolution feature map. The local and global information of the image is extracted through convolutional layers or Transformer networks.

[0055] Pixel Decoder: Next, the low-resolution feature map is upsampled through the pixel decoder to generate high-resolution embeddings for each pixel point, retaining the detailed information of the image.

[0056] Transformer Decoder: Multiple decoder layers are adopted to further optimize the image features, helping the model to identify more complex image structures and ultimately achieving accurate image segmentation.

[0057] 2DP-Specific Transfer Learning Strategy: To address the scarcity of 2DP labeled data, a three-stage transfer learning path is designed:

[0058] 1. Basic Pre-training: Pre-train the model on large-scale general datasets such as ImageNet to obtain the basic visual feature representation ability;

[0059] 2. Real 2DP Fine-tuning: Perform final fine-tuning on real 2DP labeled data to make the model accurately adapt to the characteristics of actual 2DP materials.

[0060] 3. 2DP Feature-Weighted EMA Technique: For the key features of 2DP layer detection, a feature-weighted EMA mechanism is innovatively designed:

[0061] a. Assign higher EMA weights to the key feature channels (RGB information) for the model to identify 2DP layers, enhancing the stability of these features;

[0062] b. Introduce a dynamic EMA weight adjustment strategy. According to the training results, dynamically adjust the weights of the key feature channels and adaptively adjust the EMA parameters according to the training progress and validation performance;

[0063] c. Combine model uncertainty estimation and apply higher EMA weights to the high-certainty regions (mask parts) confirmed by the Mask2Former model to further improve the model stability.

[0064] Based on the model design of this step, the present invention first proposes the MaskEMA-2DP model specifically for 2DP layer detection, which has made breakthroughs in solving the following technical difficulties:

[0065] 1. 2DP Optical Property Perception Mechanism: For the unique optical interference pattern of 2DP materials, a special optical feature extraction process (2DP-specific denoising and multi-scale contrast enhancement processes in preprocessing) is designed, which can effectively identify and enhance the interference fringe and color change features of different-layer 2DP;

[0066] 2. Multi-scale Feature Fusion Architecture: An innovative hierarchical feature fusion mechanism is designed, which combines EMA and transfer learning to achieve effective integration of multi-scale features of 2DP materials;

[0067] 3. 2DP Sample Efficient Learning Strategy: To address the scarcity of 2DP labeled data, transfer learning and EMA techniques are integrated to significantly improve the learning efficiency and stability of the model under small-sample conditions.

[0068] The MaskEMA-2DP model of the present invention is based on the Mask2Former architecture but is specifically optimized for 2DP materials. It can effectively capture layer-related features in images, thereby achieving precise segmentation of different layer regions in 2DP materials.

[0069] 2DP-specific optimization of transfer learning and EMA technology: To address the issues of insufficient 2DP labeled data and training stability, the present invention designs a collaborative optimization mechanism of transfer learning and EMA for 2DP characteristics:

[0070] 1. 2DP-specific pre-training strategy: First, pre-train the model on a general dataset to obtain basic feature representation capabilities, and finally fine-tune it on real 2DP data to form an efficient three-stage transfer learning path;

[0071] 2. 2DP feature-weighted EMA: Innovatively propose a weighted EMA technology for 2DP key features, assign higher smoothing weights to the feature channels crucial for layer recognition, and improve the stability of the model for key features;

[0072] 3. Dynamic batch normalization calibration: Design a dynamic batch normalization parameter adjustment mechanism for the variability between 2DP samples, enabling the model to better adapt to the characteristic changes of different batches of 2DP samples.

[0073] These technologies specifically optimized for 2DP significantly reduce the need for large-scale labeled data, enable the model to quickly learn the layer features of 2DP under small sample conditions, and maintain the stability of the training process.

[0074] 4. Image segmentation and layer detection

[0075] As Figure 3 shown, through the trained MaskEMA-2DP model, input optical microscope images for prediction, demonstrating the performance of the Mask2Former model in different layer detection tasks. After the image is segmented, layer classification is performed based on the pixel points of different segmented regions. Each region is classified into different layers according to the color label (such as single layer, double layer, etc.), and quantitative analysis is carried out on the percentage of each type of region. Through this process, quantitative support can be provided for the performance optimization of 2DP materials.

[0076] As Figure 3 shown in a)-d) of

[0077] As Figure 3As shown in e)-f) in [ ], statistical data of the segmentation results are given, including the area ratios of regions with different numbers of layers (for example, the single-layer accounts for 15.32%, the double-layer accounts for 52.88%, etc.). This kind of quantitative analysis helps to understand the layer distribution of two-dimensional materials under different experimental conditions and supports experimental optimization.

[0078] Model performance evaluation:

[0079] As Figure 4 shown in a)-c) in [ ], a) is the original image, b) is the result predicted by mask2former, and c) is the result predicted by the model after using EMA. The results in the figure show that EMA enhances the model's prediction ability for impurities.

[0080] As Figure 4 shown in d) in [ ], it demonstrates the improvement of the model performance by the EMA technique. Without using EMA, the mIoU (Mean Intersection over Union) value during the model training process is low. After using EMA, the mIoU value of the model increases significantly. The results in the figure show that the EMA technique enhances the model's ability to capture details and improves the accuracy of layer detection.

[0081] As Figure 4 shown in e) in [ ], it demonstrates the training efficiency of different models. After adding transfer learning, the Mask2Former model requires fewer training iterations compared to other models (such as U-Net, PSPNet, etc.), showing its efficient training performance.

[0082] 5. Experimental verification and robustness analysis

[0083] As Figure 5 shown in [ ], it demonstrates the robustness of the Mask2Former model under different brightness and contrast conditions. Experiments show that although the image brightness and contrast change, the model can still maintain a high detection accuracy.

[0084] As Figure 5 shown in a)-e) in [ ], by changing the brightness and contrast of the image, the performance of the model shows good stability. Under different processing conditions, the model can still accurately segment regions with different numbers of layers.

[0085] As Figure 5 shown in f) in [ ], it demonstrates the prediction accuracy of the model under different conditions, indicating that the model can still maintain a high accuracy when dealing with changes in brightness and contrast, verifying its wide applicability in practical applications.

[0086] As Figure 5As shown in g) in [reference], it shows the model evaluation under different image processing conditions (such as a 10% increase in brightness, a 10% decrease in contrast, etc.). The results show that even under different image variations, the Mask2Former model can still accurately identify the number of layers, and maintain a relatively high mIoU and prediction accuracy under all conditions.

[0087] 6. Final Results and Optimization

[0088] After training, the model can automatically identify the number of layers in the two-dimensional material image and perform quantitative analysis, greatly improving the efficiency of data analysis. Finally, the segmentation results and layer ratio output by the model provide data support for material optimization.

[0089] Corresponding to the foregoing embodiment of a method for detecting the number of layers of two-dimensional materials under a microscope based on deep learning, the present invention also provides an embodiment of a device for detecting the number of layers of two-dimensional materials under a microscope based on deep learning.

[0090] See Figure 6 , an embodiment of a device for detecting the number of layers of two-dimensional materials under a microscope based on deep learning provided by the embodiment of the present invention includes a memory and one or more processors. An executable code is stored in the memory. When the processor executes the executable code, it is used to implement the method for detecting the number of layers of two-dimensional materials under a microscope in the above embodiment.

[0091] The embodiment of the device for detecting the number of layers of two-dimensional materials under a microscope based on deep learning provided by the present invention can be applied to any device with data processing capabilities. The any device with data processing capabilities can be a device or apparatus such as a computer. The device embodiment can be implemented by software, or by hardware or a combination of software and hardware. Taking software implementation as an example, as a logically meaningful device, it is formed by the processor of any device with data processing capabilities reading the corresponding computer program instructions in the non-volatile memory into the memory for operation. From a hardware perspective, as Figure 6 shown, it is a hardware structure diagram of any device with data processing capabilities where the device for detecting the number of layers of two-dimensional materials under a microscope provided by the present invention is located. Except for Figure 6 the processor, memory, network interface, and non-volatile memory shown, the any device with data processing capabilities where the device in the embodiment is located usually includes other hardware according to the actual functions of the any device with data processing capabilities, which will not be elaborated here.

[0092] The implementation processes of the functions and roles of each unit in the above device are specifically detailed in the implementation processes of the corresponding steps in the above method, which will not be elaborated here.

[0093] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the descriptions in the method embodiments. The device embodiments described above are only illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of the present invention. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0094] An embodiment of the present invention also provides a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, it implements a method for detecting the number of layers of two-dimensional materials under a microscope based on deep learning in the above embodiments.

[0095] The computer-readable storage medium may be an internal storage unit of any device with data processing capabilities described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium may also be an external storage device of any device with data processing capabilities, such as a plug-in hard disk, a smart media card (SMC), an SD card, a flash card, etc. equipped on the device. Further, the computer-readable storage medium may also include both an internal storage unit and an external storage device of any device with data processing capabilities. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capabilities, and can also be used to temporarily store the data that has been output or will be output.

[0096] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the method for detecting the number of layers of two-dimensional materials under a microscope based on deep learning.

[0097] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solution of the present invention, and these simple modifications all belong to the protection scope of the present invention.

[0098] In addition, it should be noted that, in the above specific embodiments, the various specific technical features described can be combined in any suitable way without conflict. To avoid unnecessary repetition, the present invention does not separately describe various possible combination methods.

[0099] In addition, any combination can be made among various different embodiments of the present invention, as long as it does not violate the idea of the present invention, and it should equally be regarded as the content disclosed by the present invention.

Claims

1. A method for detecting the number of layers of two-dimensional materials under a microscope based on deep learning, characterized in that, For detecting the number of layers of two-dimensional polymer 2DP under a microscope, the method includes the following steps: a) Sample preparation: Mix the two-dimensional material with a solvent. Under the action of the solvent, the two-dimensional material expands and is layered under the action of shear force to obtain several layers of two-dimensional material, and deposit the two-dimensional material on a substrate; b) Image acquisition: Take a high-resolution RGB image of the deposited two-dimensional material through an optical microscope, and use it for training of a deep learning model after preprocessing; the image preprocessing includes image denoising, multi-scale contrast enhancement, brightness adaptive correction, data normalization, and data annotation; the denoising is a denoising method that combines non-local mean filtering and wavelet transform to suppress the interference patterns in the two-dimensional material image while retaining the key information related to the number of layers; the multi-scale contrast enhancement uses multi-scale adaptive histogram equalization technology to locally optimize the feature differences in different layer regions of the two-dimensional material, making the layer boundaries clearer; the brightness adaptive correction is based on the optical characteristics of the two-dimensional material and uses an adaptive brightness correction algorithm; the data annotation uses SAM to assist in annotating the two-dimensional material image to obtain the segmentation results of different layer regions; c) Model training: Based on the pre-trained Mask2Former model, after transfer learning, it is used for training of two-dimensional material images, and the exponential moving average EMA technology is used to smooth the parameters of the model; for the key features of two-dimensional material layer number detection, the following feature weighting mechanism is adopted: 1) Assign a higher EMA weight to the RGB information feature channels of the number of layers of the two-dimensional material; 2) Dynamically optimize and adjust the weights of the RGB information feature channels according to the training results; 3) Assign a higher EMA weight to the high-certainty regions confirmed by the Mask2Former model; d) Layer number detection: Apply the trained Mask2Former model to newly acquired two-dimensional material images for segmentation, generate segmentation masks for each region, and predict the number of layers of each region according to the masks; e) Layer quantification analysis: Calculate the area ratio of each layer region according to the pixel point ratio to obtain the percentage parameter of each layer.

2. The method according to claim 1, wherein The Mask2Former framework is based on the Transformer architecture and performs image segmentation through multiple layers of decoders to process the global and local features in two-dimensional material images.

3. The method according to claim 1, wherein The layer number prediction result of the Mask2Former model is displayed in the image in a color-coded manner, and different layer regions are marked with different colors.

4. The method according to claim 1, characterized in that, The layer quantification analysis outputs the percentages of each layer region and shows the distribution of different layers through statistical charts.

5. A two-dimensional material layer number detection device under a microscope based on deep learning, comprising a memory and one or more processors, wherein executable code is stored in the memory, and is characterized in that, When the processor executes the executable code, it implements a method for detecting the number of layers of two-dimensional materials under a microscope based on deep learning as described in any one of claims 1-4.

6. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements a method for detecting the number of layers of two-dimensional materials under a microscope based on deep learning as described in any one of claims 1-4.

Images