CN111709911A - Ovarian follicle automatic counting method based on neural network - Google Patents
Ovarian follicle automatic counting method based on neural network Download PDFInfo
- Publication number
- CN111709911A CN111709911A CN202010418981.7A CN202010418981A CN111709911A CN 111709911 A CN111709911 A CN 111709911A CN 202010418981 A CN202010418981 A CN 202010418981A CN 111709911 A CN111709911 A CN 111709911A
- Authority
- CN
- China
- Prior art keywords
- neural network
- follicles
- network
- model
- training
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- 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
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
- G06V10/507—Summing image-intensity values; Histogram projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20036—Morphological image processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Probability & Statistics with Applications (AREA)
- Multimedia (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses an ovarian follicle automatic counting method based on a neural network, which generates different training sets, verification sets and test sets for each time by adopting a random seed mode for all data images; continuously updating the network parameters through training of the neural network, and verifying through the IOU of the verification set so as to keep the optimal parameters of the network model; carrying out thresholding processing and noise processing on a prediction image output by a neural network to remove noise in the prediction image and convert an image into a gray image; then separating the follicles which are contacted with each other by using a distance conversion algorithm and a watershed algorithm; and finally, counting the number of the ovarian follicles by using a connected region analysis method to realize the counting function of the ovarian follicles.
Description
Technical Field
The invention relates to the field of deep learning of Watershed algorithm (Watershed) and CNN (rational Neural network), in particular to a method for automatically counting ovarian follicles of mice.
Background
The ovary is a complex endocrine organ, is an important gonadal organ for female reproduction and mainly has the functions of ovulation and female hormone secretion. Among them, the follicle plays a crucial part in the ovary, and the follicle is composed of an oocyte and many small follicular cells. The follicles in the ovary can be divided into six types according to the changes in morphology and function during follicle development: primordial follicles, primary follicles, secondary follicles, pre-antral follicles, and tertiary follicles. In the absence of any drug, normal women can find in the ovary that small follicles grow gradually to grow up and develop into follicles in different stages in one month, and a mature follicle is discharged in the ovulation phase. However, genetic mutations, toxins and some specific drugs have an effect on follicles. This is important in order to determine whether these effects are promoting or inhibiting for ovulation of the follicles.
The convolutional neural network is composed of a plurality of convolutional layers and fully-connected layers (corresponding to a classical neural network), and also comprises an activation layer and a pooling layer. Compared with other deep learning structures, the convolutional neural network can give better results in the aspects of image segmentation and recognition.
The watershed algorithm is an image segmentation algorithm based on analysis of geographic morphology, and simulates geographic structures (such as mountains, ravines, basins) to segment different objects.
Due to this importance, current stage is mainly directed to microscopic images of mouse ovary. The data set was obtained by staining and labeling the ovaries of the mice, and then by studying the development of follicles in the reproductive system of the mice and determining the number of their various follicle types.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a neural network-based ovarian follicle automatic counting method.
In order to solve the problem, the invention is realized by the following technical scheme:
a follicle counting method based on a neural network and a watershed algorithm comprises the following steps:
the method comprises the following steps: and dividing the data set into a training set, a verification set and a test set by using all data in a random seed mode. The data input by the neural network each time comprises an original picture and a marked picture;
step two: processing the pre-input pictures by using a data enhancement method for a training set when a network model is trained, and converting the picture data into a vector form;
step three: loading parameters of a pre-training model to initialize a neural network, executing forward propagation to calculate network parameters, and obtaining an output vectorO iObtaining a classification result vector y of each pixel point through a Log _ Softmax functioni(ii) a This vector yiThe maximum value of the medium weight is the final classification result predicted by the network for each pixel point, and the specific formula is yi=log_softmax(O i) ); the obtained classification result yiAnd the current correct tag valueRespectively serving as two inputs of an NLLloss loss function, and calculating a loss value; transmitting the error signal to the output of each layer, and obtaining the gradient of the network parameter through the derivative of the function of each layer to the network parameter; and updating and calculating network parameters influencing model training and model output by a stochastic gradient descent method optimizer to enable the network parameters to approach or reach an optimal value, so that a loss function is minimized or maximized.
Step four: after each training, calculating the cross-over ratio of the verification setAnd storing the maximum model parameter of IOU when the training times are up toStopping training at a certain number of times, and loading the model parameter with the largest IOU to obtain the final model of the neural network, whereinIs a true value, and y is a predicted value;
step five: the model obtained in the fourth step is an optimal model, input data are transmitted into a trained neural network to obtain an output result, the output image format is an RGB image, the RGB image is converted into a gray image through graying, an image gray histogram is established to select threshold values capable of partitioning different follicles, and then the follicles are partitioned according to the different threshold values;
step six: respectively processing the noise of the pictures obtained in the step five by using an on operationFor eliminating small objects, using closed-loop operationsFilling small cavities in the object, whereinAndrespectively, expansion and corrosion;
step seven: by distance conversionCalculating the distance from the value of each pixel to the nearest background pixel to obtain Euclidean distance map, and then finding the final erosion point of each local area, namely the closer to the center of the follicle, the larger the corresponding value, wherein X is the target point, B isxIs the closest background point to X;
step eight: then, taking the maximum value of each local area as a water injection point of a watershed, and expanding the area of each mark point from the points as much as possible, namely, either until the edge of the local area reaches or the edge of the area of another mark point reaches, thereby segmenting the follicles which are contacted with each other;
step nine: setting different thresholds for different types of follicles, judging the follicles to be noise when the area in the picture is smaller than a given threshold, analyzing and calculating the number of the connected regions in the picture by adopting the connected regions of the four neighborhoods, and marking the number of the connected regions, wherein the number of the connected regions is the number of the follicles.
Preferably, the marking picture marks different follicle types through different colors.
Preferably, the data enhancement method comprises random horizontal flipping, random vertical flipping, 360-degree random rotation, contrast, brightness, saturation, sharpening and standardization.
Compared with the prior art, the invention has the following effects:
1) the invention provides a novel ovarian follicle counting mode, which counts ovarian follicles by combining a convolution neural network of semantic segmentation and watershed calculation for the first time.
2) The number counted by the counting mode is closer to the real data, and the accuracy is higher.
Drawings
FIG. 1 is an overall flow diagram of the present invention;
fig. 2 is a diagram of a neural network architecture employed.
Detailed Description
The invention is further described with reference to the following detailed description and accompanying drawings:
fig. 1 shows a novel method for ovarian follicle counting based on neural network and watershed algorithm:
1) and dividing the data set into a training set, a verification set and a test set by using all data in a random seed mode. The data input by the neural network each time comprises an original picture and a marked picture, wherein the marked picture marks different follicle types through different colors.
2) When a network model is trained, processing a pre-input picture by using a data enhancement method (random horizontal turning, random vertical turning, 360-degree random rotation, contrast, brightness, saturation, sharpening and standardization) on a training set, and converting picture data into a vector form;
3) loading parameters of the pre-trained model to initialize the neural network (as shown in FIG. 2), performing forward propagation to calculate network parameters, and obtaining the output vectorO iObtaining a classification result vector y of each pixel point through a Log _ Softmax functioni(ii) a This vector yiThe maximum value of the medium weight is the final classification result predicted by the network for each pixel point, and the specific formula is yi=log_softmax(O i) ); the obtained classification result yiAnd the current correct tag valueRespectively serving as two inputs of an NLLloss loss function, and calculating a loss value; transmitting the error signal to the output of each layer, and obtaining the gradient of the network parameter through the derivative of the function of each layer to the network parameter; and updating and calculating network parameters influencing model training and model output by a stochastic gradient descent method optimizer to enable the network parameters to approach or reach an optimal value, so that a loss function is minimized or maximized.
4) After each training, calculating the cross-over ratio of the verification setStoring the maximum model parameter of IOU, stopping training when the training times reach a certain number, and loading the maximum model parameter of IOU to obtain the final model of the neural network, whereinIs a true value, and y is a predicted value;
5) the model obtained in the fourth step is an optimal model, input data are transmitted into a trained neural network to obtain an output result, the output image format is an RGB image, the RGB image is converted into a gray image through graying, an image gray histogram is established to select threshold values capable of partitioning different follicles, and then the follicles are partitioned according to the different threshold values;
6) respectively processing the noise of the pictures obtained in the step five by using an on operationFor eliminating small objects, using closed-loop operationsFilling small cavities in the object, whereinAndrespectively, expansion and corrosion;
7) by distance conversionCalculating the distance from the value of each pixel to the nearest background pixel to obtain Euclidean distance map, and then finding the final erosion point of each local area, namely the closer to the center of the follicle, the larger the corresponding value, wherein X is the target point, B isxIs the closest background point to X;
8) then, taking the maximum value of each local area as a water injection point of a watershed, and expanding the area of each mark point from the points as much as possible, namely, either until the edge of the local area reaches or the edge of the area of another mark point reaches, thereby segmenting the follicles which are contacted with each other;
9) setting different thresholds for different types of follicles, judging the follicles to be noise when the area in the picture is smaller than a given threshold, analyzing and calculating the number of the connected regions in the picture by adopting the connected regions of the four neighborhoods, and marking the number of the connected regions, wherein the number of the connected regions is the number of the follicles.
Claims (3)
1. A method for automatically counting ovarian follicles based on a neural network is characterized by comprising the following steps:
the method comprises the following steps: dividing a data set into a training set, a verification set and a test set by using all data in a random seed mode; the data input by the neural network each time comprises an original picture and a marked picture;
step two: processing the pre-input pictures by using a data enhancement method for a training set when a network model is trained, and converting the picture data into a vector form;
step three: loading parameters of a pre-training model to initialize a neural network, executing forward propagation to calculate network parameters, and obtaining an output vectorO iObtaining a classification result vector y of each pixel point through a Log _ Softmax functioni(ii) a This vector yiThe maximum value of the medium weight is the final classification result predicted by the network for each pixel point, and the specific formula is yi=log_soft max(O i) ); the obtained classification result yiAnd the current correct tag valueRespectively serving as two inputs of an NLLloss loss function, and calculating a loss value; transmitting the error signal to the output of each layer, and obtaining the gradient of the network parameter through the derivative of the function of each layer to the network parameter; updating and calculating network parameters influencing model training and model output by a random gradient descent method optimizer to enable the network parameters to approach or reach an optimal value, so that a loss function is minimized or maximized;
step four: after each training, calculating the cross-over ratio of the verification setStoring the maximum model parameter of IOU, stopping training when the training times reach a certain number, and loading the maximum model parameter of IOU to obtain the final model of the neural network, whereinIs a true value, and y is a predicted value;
step five: the model obtained in the fourth step is an optimal model, input data are transmitted into a trained neural network to obtain an output result, the output image format is an RGB image, the RGB image is converted into a gray image through graying, an image gray histogram is established to select threshold values capable of partitioning different follicles, and then the follicles are partitioned according to the different threshold values;
step six: respectively processing the noise of the pictures obtained in the step five by using an on operationFor eliminating small objects, using closed-loop operationsFilling small cavities in the object, whereinAndrespectively, expansion and corrosion;
step seven: by distance conversionCalculating the distance from the value of each pixel to the nearest background pixel to obtain Euclidean distance map, and then finding the final erosion point of each local area, namely the closer to the center of the follicle, the larger the corresponding value, wherein X is the target point, B isxIs the closest background point to X;
step eight: then, taking the maximum value of each local area as a water injection point of a watershed, and expanding the area of each mark point from the points as much as possible, namely, either until the edge of the local area reaches or the edge of the area of another mark point reaches, thereby segmenting the follicles which are contacted with each other;
step nine: setting different thresholds for different types of follicles, judging the follicles to be noise when the area in the picture is smaller than a given threshold, analyzing and calculating the number of the connected regions in the picture by adopting the connected regions of the four neighborhoods, and marking the number of the connected regions, wherein the number of the connected regions is the number of the follicles.
2. The method for neural network-based ovarian follicle auto-counting according to claim 1, characterized in that: the marking picture marks different follicle types through different colors.
3. The method for neural network-based ovarian follicle auto-counting according to claim 1, characterized in that: the data enhancement method comprises random horizontal turning, random vertical turning, 360-degree random rotation, contrast, brightness, saturation, sharpening and standardization.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010418981.7A CN111709911B (en) | 2020-05-18 | 2020-05-18 | Automatic ovarian follicle counting method based on neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010418981.7A CN111709911B (en) | 2020-05-18 | 2020-05-18 | Automatic ovarian follicle counting method based on neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111709911A true CN111709911A (en) | 2020-09-25 |
CN111709911B CN111709911B (en) | 2023-05-05 |
Family
ID=72538005
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010418981.7A Active CN111709911B (en) | 2020-05-18 | 2020-05-18 | Automatic ovarian follicle counting method based on neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111709911B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113744184A (en) * | 2021-07-27 | 2021-12-03 | 江苏农林职业技术学院 | Snakehead ovum counting method based on image processing |
CN114563869A (en) * | 2022-01-17 | 2022-05-31 | 中国地质大学(武汉) | Surface mount type mobile phone microscope detection system and microscopic result obtaining method thereof |
CN117982214A (en) * | 2024-04-07 | 2024-05-07 | 安徽医科大学第一附属医院 | Auxiliary identification method and system for rapid egg picking |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108416755A (en) * | 2018-03-20 | 2018-08-17 | 南昌航空大学 | A kind of image de-noising method and system based on deep learning |
CN109584251A (en) * | 2018-12-06 | 2019-04-05 | 湘潭大学 | A kind of tongue body image partition method based on single goal region segmentation |
US20200043171A1 (en) * | 2018-07-31 | 2020-02-06 | Element Ai Inc. | Counting objects in images based on approximate locations |
-
2020
- 2020-05-18 CN CN202010418981.7A patent/CN111709911B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108416755A (en) * | 2018-03-20 | 2018-08-17 | 南昌航空大学 | A kind of image de-noising method and system based on deep learning |
US20200043171A1 (en) * | 2018-07-31 | 2020-02-06 | Element Ai Inc. | Counting objects in images based on approximate locations |
CN109584251A (en) * | 2018-12-06 | 2019-04-05 | 湘潭大学 | A kind of tongue body image partition method based on single goal region segmentation |
Non-Patent Citations (2)
Title |
---|
田佳鹭;邓立国;: "基于改进VGG16的猴子图像分类方法" * |
黄斌;卢金金;王建华;吴星明;陈伟海;: "基于深度卷积神经网络的物体识别算法" * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113744184A (en) * | 2021-07-27 | 2021-12-03 | 江苏农林职业技术学院 | Snakehead ovum counting method based on image processing |
CN114563869A (en) * | 2022-01-17 | 2022-05-31 | 中国地质大学(武汉) | Surface mount type mobile phone microscope detection system and microscopic result obtaining method thereof |
CN117982214A (en) * | 2024-04-07 | 2024-05-07 | 安徽医科大学第一附属医院 | Auxiliary identification method and system for rapid egg picking |
Also Published As
Publication number | Publication date |
---|---|
CN111709911B (en) | 2023-05-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109800778B (en) | Faster RCNN target detection method based on difficultly-divided sample mining | |
CN111325152B (en) | Traffic sign recognition method based on deep learning | |
CN111898406B (en) | Face detection method based on focus loss and multitask cascade | |
CN111709911A (en) | Ovarian follicle automatic counting method based on neural network | |
CN108805070A (en) | A kind of deep learning pedestrian detection method based on built-in terminal | |
CN110532946B (en) | Method for identifying axle type of green-traffic vehicle based on convolutional neural network | |
CN112287941B (en) | License plate recognition method based on automatic character region perception | |
CN113096096B (en) | Microscopic image bone marrow cell counting method and system fusing morphological characteristics | |
CN111833322B (en) | Garbage multi-target detection method based on improved YOLOv3 | |
CN111986126B (en) | Multi-target detection method based on improved VGG16 network | |
CN111986125A (en) | Method for multi-target task instance segmentation | |
CN108537168A (en) | Human facial expression recognition method based on transfer learning technology | |
CN113420643A (en) | Lightweight underwater target detection method based on depth separable cavity convolution | |
CN111127360A (en) | Gray level image transfer learning method based on automatic encoder | |
CN111931572B (en) | Target detection method for remote sensing image | |
CN117788957B (en) | Deep learning-based qualification image classification method and system | |
CN113408524A (en) | Crop image segmentation and extraction algorithm based on MASK RCNN | |
CN107871315B (en) | Video image motion detection method and device | |
CN113221956A (en) | Target identification method and device based on improved multi-scale depth model | |
CN114550134A (en) | Deep learning-based traffic sign detection and identification method | |
CN114140485A (en) | Method and system for generating cutting track of main root of panax notoginseng | |
CN117765480A (en) | Method and system for early warning migration of wild animals along road | |
CN112926694A (en) | Method for automatically identifying pigs in image based on improved neural network | |
CN112132207A (en) | Target detection neural network construction method based on multi-branch feature mapping | |
CN113837236B (en) | Method and device for identifying target object in image, terminal equipment and storage medium |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |