CN111860619A - Industrial detection AI intelligent model for deep learning - Google Patents
Industrial detection AI intelligent model for deep learning Download PDFInfo
- Publication number
- CN111860619A CN111860619A CN202010627346.XA CN202010627346A CN111860619A CN 111860619 A CN111860619 A CN 111860619A CN 202010627346 A CN202010627346 A CN 202010627346A CN 111860619 A CN111860619 A CN 111860619A
- Authority
- CN
- China
- Prior art keywords
- module
- layer
- learning
- network
- different
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of industrial detection, in particular to an AI intelligent model for deep learning for industrial detection, which can greatly reduce the number of network learning parameters and obtain a more simplified network structure; meanwhile, the requirement for a large amount of data of a general deep learning model is reduced in the training process, and the method is more suitable for automatic defect detection of a production line with difficultly-obtained training data; it is characterized by comprising: the system comprises a space multi-scale module, a channel feature re-dynamic calibration module, a feature reuse module of different levels of pyramids, a global connection pooling operation module and a batch standardization module.
Description
Technical Field
The invention relates to the technical field of industrial detection, in particular to an AI intelligent model for deep learning for industrial detection.
Background
In image recognition, deep learning is widely used, and particularly, in a data-based classification application or a learning application called a supervised learning application, a data model is generated according to a classification result of data in advance on the premise that a data classification result is known, and a predefined model classification result is obtained by adjusting parameters and weight values of the model. What is called deep learning is that the model is divided into many layers, and the number of layers of the extreme model is even more than 1000. For each layer, specific neurons are connected with specific weights and parameters, while neurons between each layer of a fully connected layer are all connected, making the parameters extremely large. To solve the problem that the input image has a pixel value of 1000X1000 and the weights of the upper layer and the lower layer reach 10^12, the convolutional neural network limits the direct connection of the upper layer neuron and the lower layer neuron, and the weights are calculated through a weight sharing and pooling strategy by taking a convolutional kernel as an intermediary. Convolutional networks are a multi-layered perceptron specifically designed to recognize two-dimensional shapes, the structure of which is highly invariant to translation, scaling, tilting, or other forms of deformation. Its advantages include less parameters to be trained, low spatial resolution of network, and no small offset and distortion of signal, so having low requirement on translation invariance of input data.
The deep learning algorithm based on the convolutional neural network is widely applied to a plurality of aspects including target recognition, image segmentation, image generation and the like, such as handwritten character recognition and face recognition, and good economic benefits are obtained.
The convolutional neural network realizes the weight sharing pooling operation through the convolutional kernel, improves the generalization effect of the model, and improves the tolerance of the model to the deviation of input data, such as irregularity of handwritten characters and partial shielding and deformation of human faces, thereby having good adaptability.
However, in the field of industrial inspection, the main purpose is to identify defects of finished products and semi-finished products, on one hand, the defect detection method has strong adaptability to a strong background, that is, higher model generalization capability, and on the other hand, the defect detection method needs to have identification capability for specific defects, that is, the deviation of specific data cannot be tolerated, which is a challenge for mainstream convolutional neural networks, and requires higher data requirement for deep learning model training.
Disclosure of Invention
In order to solve the technical problems, the invention provides a network learning method which can greatly reduce the number of network learning parameters and obtain a more simplified network structure; meanwhile, the requirement for a large amount of data of a general deep learning model is reduced in the training process, and the deep learning AI intelligent model is more suitable for automatic defect detection of the production line, the training data of which are difficult to obtain, and is used for industrial detection.
The invention relates to an AI intelligent model for deep learning for industrial detection, which comprises:
the system comprises a space multi-scale module, a channel feature re-dynamic calibration module, a feature reuse module of different levels of pyramids, a global connection pooling operation module and a batch standardization module;
the spatial multi-scale module is used for simulating a learning and reasoning mode combining coarse granularity and fine granularity of human beings based on a spatial multi-scale theory, embedding multi-scale information in the structure, aggregating the characteristics on various different receptive fields to obtain performance gain, and comprehensively considering the large context characteristics and the local fine and fine characteristics of the surrounding environment of the image;
the channel characteristic re-dynamic calibration module is used for dynamically re-calibrating or weighting different channels for different input images according to different channel characteristic information with different importance of the final application task;
the characteristic reusing module of the pyramid with different levels reuses the characteristics of the front layer of the network, reduces the learned network parameters, if the characteristics of the front layer are finally classified or identified, the back network layer does not need to use a more complex network structure to relearn the front characteristics, and can reduce the corresponding data volume for training;
The global connection pooling operation module only uses one full connection layer, greatly reduces the number of parameters compared with the conventional three-layer full connection layer, utilizes a global average pooling layer of the feature map before the classification layer for classification tasks, enhances the learning of feature map categories, and is less prone to overfitting compared with the conventional multi-layer full connection layer;
and the batch standardization module is flexibly configured to add batch standardization layers among one or more convolution layers.
The invention relates to an AI intelligent model for deep learning for industrial detection, which further comprises:
the spatial multi-scale module can also effectively extract complementary information by adopting parallel multi-scale information channels when the spatial multi-scale module is close to an input image layer.
The invention relates to an AI intelligent model for deep learning for industrial detection, which further comprises:
the channel characteristic re-dynamic calibration module can also adaptively re-calibrate the characteristics of each channel by learning the correlation between channels of each layer by using global information.
The invention relates to an AI intelligent model for deep learning for industrial detection, which further comprises:
the characteristic reusing modules of the pyramids in different levels can flexibly configure the through connection between the network layers according to the requirement.
Compared with the prior art, the invention has the beneficial effects that: the user can flexibly customize according to the application; flexible network architecture, flexibly configured by a variety of different basic modules; the novel network architecture greatly reduces the number of network learning parameters and obtains a more simplified network structure; the requirement for a large amount of data of a general deep learning model is reduced in the training process, and the method is more suitable for automatic defect detection of a production line with difficult acquisition of training data; under the condition of the same training data volume, the probability of overfitting is greatly reduced, so that the generalization capability of the learning model is improved; the method reduces the operation amount, is suitable for the rapid detection of a real-time system, and is easy to integrate in an embedded system; and aiming at the directional reinforcement of specific characteristics, the method has strong identification capability on specified defects.
Drawings
FIG. 1 is a logical relationship diagram of the present invention.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
An AI intelligent model for deep learning for industrial inspection, comprising: the system comprises a space multi-scale module, a channel feature re-dynamic calibration module, a feature reuse module of different levels of pyramids, a global connection pooling operation module and a batch standardization module; the spatial multi-scale module is used for simulating a learning and reasoning mode combining coarse granularity and fine granularity of human beings based on a spatial multi-scale theory, embedding multi-scale information in the structure, aggregating the characteristics on various different receptive fields to obtain performance gain, and comprehensively considering the large context characteristics and the local fine and fine characteristics of the surrounding environment of the image; the channel characteristic re-dynamic calibration module is used for dynamically re-calibrating or weighting different channels for different input images according to different channel characteristic information with different importance of the final application task; the characteristic reusing module of the pyramid with different levels reuses the characteristics of the front layer of the network, reduces the learned network parameters, if the characteristics of the front layer are finally classified or identified, the back network layer does not need to use a more complex network structure to relearn the front characteristics, can reduce the corresponding data volume for training, greatly reduces the parameter quantity of the network structure, has extremely high parameter utilization rate, and effectively solves the learning difficulty caused by the phenomena of gradient and information disappearance; the global connection pooling operation module only uses one full connection layer, greatly reduces the number of parameters compared with the conventional three-layer full connection layer, utilizes a global average pooling layer of the feature map before the classification layer for classification tasks, enhances the learning of feature map categories, and is less prone to overfitting compared with the conventional multi-layer full connection layer; the batch standardization module is flexibly configured between one or more convolution layers and added with batch standardization layers, so that the hyper-parameters are easy to optimize, and the learning efficiency and the convergence speed are improved; the AI model can accurately classify various defects under a complex background, even can effectively distinguish weak defects and dust phenomena, and greatly reduces excessive killing caused by environmental problems in industrial production; the user can flexibly customize according to the application; flexible network architecture, flexibly configured by a variety of different basic modules; the novel network architecture greatly reduces the number of network learning parameters and obtains a more simplified network structure; the requirement for a large amount of data of a general deep learning model is reduced in the training process, and the method is more suitable for automatic defect detection of a production line with difficult acquisition of training data; under the condition of the same training data volume, the probability of overfitting is greatly reduced, so that the generalization capability of the learning model is improved; the method reduces the operation amount, is suitable for the rapid detection of a real-time system, and is easy to integrate in an embedded system; and aiming at the directional reinforcement of specific characteristics, the method has strong identification capability on specified defects.
As a preferred technical solution, an AI intelligent model for deep learning for industrial inspection is characterized by comprising: the system comprises a space multi-scale module, a channel feature re-dynamic calibration module, a feature reuse module of different levels of pyramids, a global connection pooling operation module and a batch standardization module; the spatial multi-scale module is used for simulating a learning and reasoning mode combining coarse granularity and fine granularity of human beings based on a spatial multi-scale theory, embedding multi-scale information in the structure, aggregating the characteristics on various different receptive fields to obtain performance gain, and comprehensively considering the large context characteristics and the local fine and fine characteristics of the surrounding environment of the image; the channel characteristic re-dynamic calibration module is used for dynamically re-calibrating or weighting different channels for different input images according to different channel characteristic information with different importance of the final application task; the characteristic reusing module of the pyramid with different levels reuses the characteristics of the front layer of the network, reduces the learned network parameters, if the characteristics of the front layer are finally classified or identified, the back network layer does not need to use a more complex network structure to relearn the front characteristics, can reduce the corresponding data volume for training, greatly reduces the parameter quantity of the network structure, has extremely high parameter utilization rate, and effectively solves the learning difficulty caused by the phenomena of gradient and information disappearance; the global connection pooling operation module only uses one full connection layer, greatly reduces the number of parameters compared with the conventional three-layer full connection layer, utilizes a global average pooling layer of the feature map before the classification layer for classification tasks, enhances the learning of feature map categories, and is less prone to overfitting compared with the conventional multi-layer full connection layer; the batch standardization module is flexibly configured between one or more convolution layers and added with batch standardization layers, so that the hyper-parameters are easy to optimize, and the learning efficiency and the convergence speed are improved; the spatial multi-scale module can also effectively extract complementary information by adopting parallel multi-scale information channels when the spatial multi-scale module is close to an input image layer.
As a preferred technical solution, an AI intelligent model for deep learning for industrial inspection is characterized by comprising: the system comprises a space multi-scale module, a channel feature re-dynamic calibration module, a feature reuse module of different levels of pyramids, a global connection pooling operation module and a batch standardization module; the spatial multi-scale module is used for simulating a learning and reasoning mode combining coarse granularity and fine granularity of human beings based on a spatial multi-scale theory, embedding multi-scale information in the structure, aggregating the characteristics on various different receptive fields to obtain performance gain, and comprehensively considering the large context characteristics and the local fine and fine characteristics of the surrounding environment of the image; the channel characteristic re-dynamic calibration module is used for dynamically re-calibrating or weighting different channels for different input images according to different channel characteristic information with different importance of the final application task; the characteristic reusing module of the pyramid with different levels reuses the characteristics of the front layer of the network, reduces the learned network parameters, if the characteristics of the front layer are finally classified or identified, the back network layer does not need to use a more complex network structure to relearn the front characteristics, can reduce the corresponding data volume for training, greatly reduces the parameter quantity of the network structure, has extremely high parameter utilization rate, and effectively solves the learning difficulty caused by the phenomena of gradient and information disappearance; the global connection pooling operation module only uses one full connection layer, greatly reduces the number of parameters compared with the conventional three-layer full connection layer, utilizes a global average pooling layer of the feature map before the classification layer for classification tasks, enhances the learning of feature map categories, and is less prone to overfitting compared with the conventional multi-layer full connection layer; the batch standardization module is flexibly configured between one or more convolution layers and added with batch standardization layers, so that the hyper-parameters are easy to optimize, and the learning efficiency and the convergence speed are improved; the channel characteristic re-dynamic calibration module can also adaptively re-calibrate the characteristics of each channel by learning the correlation between channels of each layer by using global information; global information can be used by this mechanism to selectively enhance the characteristics of information that is more beneficial to the task and suppress characteristics of useless, interfering, redundant information, such as noise.
As a preferred technical solution, an AI intelligent model for deep learning for industrial inspection is characterized by comprising: the system comprises a space multi-scale module, a channel feature re-dynamic calibration module, a feature reuse module of different levels of pyramids, a global connection pooling operation module and a batch standardization module; the spatial multi-scale module is used for simulating a learning and reasoning mode combining coarse granularity and fine granularity of human beings based on a spatial multi-scale theory, embedding multi-scale information in the structure, aggregating the characteristics on various different receptive fields to obtain performance gain, and comprehensively considering the large context characteristics and the local fine and fine characteristics of the surrounding environment of the image; the channel characteristic re-dynamic calibration module is used for dynamically re-calibrating or weighting different channels for different input images according to different channel characteristic information with different importance of the final application task; the characteristic reusing module of the pyramid with different levels reuses the characteristics of the front layer of the network, reduces the learned network parameters, if the characteristics of the front layer are finally classified or identified, the back network layer does not need to use a more complex network structure to relearn the front characteristics, can reduce the corresponding data volume for training, greatly reduces the parameter quantity of the network structure, has extremely high parameter utilization rate, and effectively solves the learning difficulty caused by the phenomena of gradient and information disappearance; the global connection pooling operation module only uses one full connection layer, greatly reduces the number of parameters compared with the conventional three-layer full connection layer, utilizes a global average pooling layer of the feature map before the classification layer for classification tasks, enhances the learning of feature map categories, and is less prone to overfitting compared with the conventional multi-layer full connection layer; the batch standardization module is flexibly configured between one or more convolution layers and added with batch standardization layers, so that the hyper-parameters are easy to optimize, and the learning efficiency and the convergence speed are improved; the characteristic reusing modules of the pyramids in different levels can flexibly configure the through connection between the network layers according to the requirement; because of the direct connection of the output layer to the front layer, the learning efficiency and the convergence speed are improved.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.
Claims (4)
1. An AI intelligent model for deep learning for industrial inspection, comprising: the system comprises a space multi-scale module, a channel feature re-dynamic calibration module, a feature reuse module of different levels of pyramids, a global connection pooling operation module and a batch standardization module;
the spatial multi-scale module is used for simulating a learning and reasoning mode combining coarse granularity and fine granularity of human beings based on a spatial multi-scale theory, embedding multi-scale information in the structure, aggregating the characteristics on various different receptive fields to obtain performance gain, and comprehensively considering the large context characteristics and the local fine and fine characteristics of the surrounding environment of the image;
the channel characteristic re-dynamic calibration module is used for dynamically re-calibrating or weighting different channels for different input images according to different channel characteristic information with different importance of the final application task;
The characteristic reusing module of the pyramid with different levels reuses the characteristics of the front layer of the network, reduces the learned network parameters, if the characteristics of the front layer are finally classified or identified, the back network layer does not need to use a more complex network structure to relearn the front characteristics, and can reduce the corresponding data volume for training;
the global connection pooling operation module only uses one full connection layer, greatly reduces the number of parameters compared with the conventional three-layer full connection layer, utilizes a global average pooling layer of the feature map before the classification layer for classification tasks, enhances the learning of feature map categories, and is less prone to overfitting compared with the conventional multi-layer full connection layer;
and the batch standardization module is flexibly configured to add batch standardization layers among one or more convolution layers.
2. The AI intelligence model for deep learning for industrial testing as recited in claim 1, wherein the spatial multiscale module is further capable of efficiently extracting complementary information using parallel multiscale information channels when approaching the input image layer.
3. The deep-learning AI intelligence model for industrial testing as recited in claim 1, wherein the channel characteristic re-dynamic calibration module is further capable of adaptively re-calibrating the characteristics of each channel by learning the correlation between each layer of channels using global information.
4. The AI intelligent model for deep learning for industrial inspection according to claim 1, wherein the feature reuse modules of different levels of pyramids can flexibly configure the through connection between the network layers according to the requirement.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010627346.XA CN111860619A (en) | 2020-07-02 | 2020-07-02 | Industrial detection AI intelligent model for deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010627346.XA CN111860619A (en) | 2020-07-02 | 2020-07-02 | Industrial detection AI intelligent model for deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111860619A true CN111860619A (en) | 2020-10-30 |
Family
ID=72989084
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010627346.XA Pending CN111860619A (en) | 2020-07-02 | 2020-07-02 | Industrial detection AI intelligent model for deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111860619A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107437096A (en) * | 2017-07-28 | 2017-12-05 | 北京大学 | Image classification method based on the efficient depth residual error network model of parameter |
| CN109145920A (en) * | 2018-08-21 | 2019-01-04 | 电子科技大学 | A kind of image, semantic dividing method based on deep neural network |
| CN109376576A (en) * | 2018-08-21 | 2019-02-22 | 中国海洋大学 | The object detection method for training network from zero based on the intensive connection of alternately update |
| CN109614985A (en) * | 2018-11-06 | 2019-04-12 | 华南理工大学 | A kind of object detection method based on intensive connection features pyramid network |
| CN110110751A (en) * | 2019-03-31 | 2019-08-09 | 华南理工大学 | A kind of Chinese herbal medicine recognition methods of the pyramid network based on attention mechanism |
| CN111274942A (en) * | 2020-01-19 | 2020-06-12 | 国汽(北京)智能网联汽车研究院有限公司 | Traffic cone identification method and device based on cascade network |
-
2020
- 2020-07-02 CN CN202010627346.XA patent/CN111860619A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107437096A (en) * | 2017-07-28 | 2017-12-05 | 北京大学 | Image classification method based on the efficient depth residual error network model of parameter |
| CN109145920A (en) * | 2018-08-21 | 2019-01-04 | 电子科技大学 | A kind of image, semantic dividing method based on deep neural network |
| CN109376576A (en) * | 2018-08-21 | 2019-02-22 | 中国海洋大学 | The object detection method for training network from zero based on the intensive connection of alternately update |
| CN109614985A (en) * | 2018-11-06 | 2019-04-12 | 华南理工大学 | A kind of object detection method based on intensive connection features pyramid network |
| CN110110751A (en) * | 2019-03-31 | 2019-08-09 | 华南理工大学 | A kind of Chinese herbal medicine recognition methods of the pyramid network based on attention mechanism |
| CN111274942A (en) * | 2020-01-19 | 2020-06-12 | 国汽(北京)智能网联汽车研究院有限公司 | Traffic cone identification method and device based on cascade network |
Non-Patent Citations (2)
| Title |
|---|
| 李瑞坤: "基于多尺度特征的锂电池表面缺陷检测方法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》, pages 26 - 27 * |
| 谷静等: "基于卷积神经网络的焊缝缺陷图像分类研究", 《应用光学》, vol. 41, no. 3, pages 531 - 537 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108133188B (en) | Behavior identification method based on motion history image and convolutional neural network | |
| CN110084281B (en) | Image generation method, neural network compression method, related device and equipment | |
| Alani et al. | Hand gesture recognition using an adapted convolutional neural network with data augmentation | |
| CN113705769B (en) | Neural network training method and device | |
| CN111950649B (en) | Attention mechanism and capsule network-based low-illumination image classification method | |
| CN106845487B (en) | End-to-end license plate identification method | |
| Liang et al. | Recurrent convolutional neural network for object recognition | |
| Wang et al. | Deep learning algorithms with applications to video analytics for a smart city: A survey | |
| CA2993011C (en) | Enforced sparsity for classification | |
| CN111291809B (en) | Processing device, method and storage medium | |
| Chen et al. | Research on recognition of fly species based on improved RetinaNet and CBAM | |
| CN104537387A (en) | Method and system for classifying automobile types based on neural network | |
| Yang et al. | Semi-supervised learning of feature hierarchies for object detection in a video | |
| Li et al. | Pedestrian detection based on deep learning model | |
| CN112395951B (en) | Complex scene-oriented domain-adaptive traffic target detection and identification method | |
| CN104182772A (en) | Gesture recognition method based on deep learning | |
| CN113469236A (en) | Deep clustering image recognition system and method for self-label learning | |
| CN108416795B (en) | Video action identification method based on sorting pooling fusion space characteristics | |
| CN116363535A (en) | Ship detection method in unmanned aerial vehicle aerial image based on convolutional neural network | |
| CN117372853A (en) | Underwater target detection algorithm based on image enhancement and attention mechanism | |
| Ding et al. | Object detection method based on lightweight YOLOv4 and attention mechanism in security scenes | |
| CN108537132B (en) | Road segmentation method of depth automatic encoder based on supervised learning | |
| CN113221683A (en) | Expression recognition method based on CNN model in teaching scene | |
| CN116863418A (en) | Road traffic limiting vehicle species division method based on CA_MixNet | |
| Luo et al. | Edgeactnet: Edge intelligence-enabled human activity recognition using radar point cloud |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |