CN116229236A - Bacillus tuberculosis detection method based on improved YOLO v5 model - Google Patents
Bacillus tuberculosis detection method based on improved YOLO v5 model Download PDFInfo
- Publication number
- CN116229236A CN116229236A CN202310222513.6A CN202310222513A CN116229236A CN 116229236 A CN116229236 A CN 116229236A CN 202310222513 A CN202310222513 A CN 202310222513A CN 116229236 A CN116229236 A CN 116229236A
- Authority
- CN
- China
- Prior art keywords
- tubercle bacillus
- model
- frame
- detection
- improved
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- C—CHEMISTRY; METALLURGY
- C12—BIOCHEMISTRY; BEER; SPIRITS; WINE; VINEGAR; MICROBIOLOGY; ENZYMOLOGY; MUTATION OR GENETIC ENGINEERING
- C12Q—MEASURING OR TESTING PROCESSES INVOLVING ENZYMES, NUCLEIC ACIDS OR MICROORGANISMS; COMPOSITIONS OR TEST PAPERS THEREFOR; PROCESSES OF PREPARING SUCH COMPOSITIONS; CONDITION-RESPONSIVE CONTROL IN MICROBIOLOGICAL OR ENZYMOLOGICAL PROCESSES
- C12Q1/00—Measuring or testing processes involving enzymes, nucleic acids or microorganisms; Compositions therefor; Processes of preparing such compositions
- C12Q1/02—Measuring or testing processes involving enzymes, nucleic acids or microorganisms; Compositions therefor; Processes of preparing such compositions involving viable microorganisms
- C12Q1/04—Determining presence or kind of microorganism; Use of selective media for testing antibiotics or bacteriocides; Compositions containing a chemical indicator therefor
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/695—Preprocessing, e.g. image segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/698—Matching; Classification
-
- C—CHEMISTRY; METALLURGY
- C12—BIOCHEMISTRY; BEER; SPIRITS; WINE; VINEGAR; MICROBIOLOGY; ENZYMOLOGY; MUTATION OR GENETIC ENGINEERING
- C12R—INDEXING SCHEME ASSOCIATED WITH SUBCLASSES C12C - C12Q, RELATING TO MICROORGANISMS
- C12R2001/00—Microorganisms ; Processes using microorganisms
- C12R2001/01—Bacteria or Actinomycetales ; using bacteria or Actinomycetales
- C12R2001/32—Mycobacterium
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2333/00—Assays involving biological materials from specific organisms or of a specific nature
- G01N2333/195—Assays involving biological materials from specific organisms or of a specific nature from bacteria
- G01N2333/35—Assays involving biological materials from specific organisms or of a specific nature from bacteria from Mycobacteriaceae (F)
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The invention belongs to the technical field of target detection, and particularly relates to a tubercle bacillus detection method based on an improved YOLO v5 model, which comprises the following steps: inputting a tubercle bacillus image in a sputum sample into a trained improved YOLOv5 model; deep characteristic information is obtained through a Backbone network of a backhaul, and characteristic diagrams with different depths are obtained; obtaining tensor data of different scales after up-sampling and feature fusion in an FPNs network; and respectively predicting classification and prediction frames of the targets by using tensor data of different scales through Head heads to obtain detection results. According to the invention, the improved YOLOv5 algorithm is used for detecting the tubercle bacillus, so that the detection effect is better than that of the conventional SSD and Faster R-CNN algorithms, the tubercle bacillus can be accurately identified in the actual sputum sample detection process, the tubercle bacillus detection efficiency in the medical field is improved, and important help is provided for disease diagnosis.
Description
Technical Field
The invention belongs to the technical field of target detection, and particularly relates to a tubercle bacillus detection method based on an improved YOLO v5 model.
Background
In order to diagnose tuberculosis, a method for detecting tubercle bacillus in sputum samples is a leading-edge subject in current medical image detection under the promotion of deep learning and medical image field combined detection research. The most rapid and accurate detection method for detecting the tubercle bacillus is to manually screen, and a doctor observes the dyed tubercle bacillus through a microscope to diagnose whether a patient is a tubercle patient or not, but the detection method has some defects. First, this method relies heavily on subjective judgment of the relevant person, which is required to have abundant knowledge and experience for detecting tubercle bacillus. Secondly, the related work of detecting the tubercle bacillus is work with large workload and easy fatigue, and related personnel are easy to miss detection or miss detection. Therefore, the target detection technology based on deep learning detects the tubercle bacillus so as to diagnose the tuberculosis, has important application value and is an important development direction of medical detection.
Along with the rapid development of deep learning, image processing and the like in the field of medical image detection, the detection of the tubercle bacillus by using a target detection algorithm in computer vision has become a necessary trend of development, and the efficiency of detecting the tubercle bacillus by a clinician in a short time is greatly improved. The YOLOv5 target detection algorithm is a classical method in the target detection direction in recent years, and the YOLOv5 target detection algorithm can well complete the identification detection of small targets and is used for detecting tubercle bacillus, and good effects are achieved.
The detection of tubercle bacillus by the YOLOv5 target detection algorithm still cannot reach the accuracy of manual detection, and is not applied to actual detection. And the current YOLOv5 algorithm has the problems of insufficient position information, inaccurate identification and the like when identifying tubercle bacillus in sputum samples with complex backgrounds. Small target detection is still one of the most challenging tasks in target detection, where tubercle bacillus detection is one of the standard small target detection.
Disclosure of Invention
In order to solve the technical problems, the invention provides a tubercle bacillus detection method based on an improved YOLO v5 model, which comprises the following steps:
inputting a tubercle bacillus image in a sputum sample into a trained improved YOLOv5 model, and detecting tubercle bacillus in the image;
the improved YOLOv5 model comprises: input end, improved Backbone network, FPNs network, head;
the training process for improving the YOLOv5 model comprises the following steps:
s1: acquiring a data set of a tubercle bacillus image in a sputum sample, marking the acquired data, performing format conversion treatment, and dividing a training set and the data set from the data set after format conversion;
s2: inputting the data in the training set to an input end of an improved YOLOv5 model for preprocessing;
s3: inputting the preprocessed picture into a Backbone network of a backhaul to obtain deep characteristic information, and obtaining characteristic diagrams with different depths;
s4: inputting feature maps with different depths into an FPNs network to perform up-sampling and feature fusion to obtain tensor data with different scales;
s5: tensor data of different scales are used for respectively predicting classification and prediction frames of targets through Head heads;
s6: selecting a correct detection frame according to the prediction frame and the real frame, and eliminating redundant prediction frames by using a non-maximum suppression NMS method;
s7: and calculating a loss function of the model according to the detection result, then adjusting parameters of the model, and finishing model training when the change of the loss function value is floating or not smaller.
The invention has the beneficial effects that:
according to the invention, the improved YOLOv5 algorithm is used for detecting the tubercle bacillus, so that the detection effect is better than that of the conventional SSD and Faster R-CNN algorithms, the tubercle bacillus can be accurately identified in the actual sputum sample detection process, the tubercle bacillus detection efficiency in the medical field is improved, and important help is provided for disease diagnosis.
Drawings
FIG. 1 is a flowchart of the training process of the improved YOLOv5 model of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
A tubercle bacillus detection method based on an improved YOLO v5 model comprises the following steps:
inputting a tubercle bacillus image in a sputum sample into a trained improved YOLOv5 model, and detecting tubercle bacillus in the image;
the improved YOLOv5 model comprises: input end, improved Backbone network, FPNs network, head;
as shown in fig. 1, the training process for improving the YOLOv5 model includes the following steps:
s1: acquiring a data set of a tubercle bacillus image in a sputum sample, marking the acquired data, performing format conversion treatment, and dividing a training set and the data set from the data set after format conversion;
s2: inputting the data in the training set to an input end of an improved YOLOv5 model for preprocessing;
s3: inputting the preprocessed picture into a Backbone network of a backhaul to obtain deep characteristic information, and obtaining characteristic diagrams with different depths;
s4: inputting feature maps with different depths into an FPNs network to perform up-sampling and feature fusion to obtain tensor data with different scales;
s5: tensor data of different scales are used for respectively predicting classification and prediction frames of targets through Head heads;
s6: selecting a correct detection frame according to the prediction frame and the real frame, and eliminating redundant prediction frames by using a non-maximum suppression NMS method;
s7: and calculating a loss function of the model according to the detection result, then adjusting parameters of the model, and finishing model training when the change of the loss function value is floating or not smaller.
The improved backhaul Backbone network comprises: a Focus module and a CSPDarknet53 network;
merging a MHSA multi-head self-attention mechanism into each BottleNeck module of a C3 module of the last layer of the CSPDarknet53 network to obtain three BottleNeck modules of the last layer of the CSPDarknet53 network to become three BottleNeck Transformer modules;
the picture is divided into pixels through a Focus module of an improved CSPDarknet53 network, four matrixes which are unchanged in channel number and scaled into one fourth of the original size are obtained, the four input matrixes are then combined according to the channel number and input into the CSPDarknet53 network, and a characteristic diagram with the depth of 160 multiplied by 160, 80 multiplied by 80 and 40 multiplied by 40 is output through multi-layer convolution and MHSA multi-head self-attention mechanism operation.
The FPNs network comprises:
target fusion detection was performed using a 160×160 large-scale detection layer, an 80×80 medium-scale detection layer, and a 40×40 small-scale detection layer.
And marking the tubercle bacillus in the sputum sample data set, and storing the marked image into a VOC data set format.
The preprocessing of the data in the training set by the input end comprises the following steps:
and carrying out random horizontal overturning, color adjustment, multi-scale input, sample mixing in proportion, cutting and splicing on the input picture to obtain the image with enhanced data.
Classification of targets by Head prediction on feature maps of different scales, comprising:
conv operation is carried out on tensor data with different scales through convolution kernels with the channel number of 6 and 3 multiplied by 3 to obtain a confidence coefficient parameter of the target, a confidence coefficient threshold is set, when the obtained confidence coefficient parameter is larger than the confidence coefficient threshold, the detection target is tubercle bacillus, and otherwise, the detection target is abandoned.
Specifically, the confidence coefficient threshold may be used to represent the reliability degree of the overall confidence coefficient, and if the confidence coefficient parameter is greater than the set confidence coefficient threshold, the confidence coefficient parameter may be considered to be reliable, and the detection target may be directly obtained; if the confidence parameter is smaller than the set confidence threshold, the detection target is discarded if the confidence parameter is not reliable, the preset confidence threshold may be set to 0.8, the confidence may represent a probability of possibility, and in this example, an overall confidence median of 0.8 or more may be regarded as an accuracy of more than 80%. The confidence threshold set above may be changed flexibly according to circumstances, and is not particularly limited herein.
Regression through Head prediction bounding boxes on feature maps of different scales, i.e. prediction boxes of targets, includes:
conv operation is carried out on tensor data with different scales through convolution kernels with the channel number of 6 and 3 multiplied by 3, 4 position information parameters are obtained, and the parameters respectively represent the length and the width of a prediction frame and the vertical and horizontal distances of the prediction frame from the upper left corner of an image.
Selecting a correct detection frame according to the detection frame and the real frame, including:
and calculating IoU values of a prediction frame and a real frame generated by the model on the tubercle bacillus, taking the IoU value as a threshold value of whether the model identifies the tubercle bacillus, and when the prediction frame exceeds the IoU threshold value, considering that the correct tubercle bacillus is detected, so as to obtain a detection frame with the correct prediction frame.
IoU values of the prediction box and the real box generated by the calculation model on the tubercle bacillus are included:
wherein B represents a prediction frame, B GT Representing a real box.
Calculating a SIoU loss function of the model, comprising:
wherein L is box A SIoU loss function representing the model, ioU represents IoU values of the model for the prediction box and the true box generated by the tubercle bacillus,b represents a prediction frame, B GT Representing the true box, delta represents the angular cost,ρ t indicating angle loss->C w Representing the length of a rectangle with the line connecting the center points of the real frame and the predicted frame as diagonal lines, C h A width of a rectangle representing that a line connecting a center point of a real frame and a predicted frame is a diagonal line, γ represents an angle loss offset, γ=2- Λ, Λ represents a distance cost, +.>Omega represents shape cost, < >>ω t Representing shape loss,/>w、h、w gt 、h gt Representing the width and height of the predicted and real frames, respectively, θ represents the degree of attention to control the shape loss, and the value of θ is set to 4 herein in order to avoid too much attention to the shape loss and reduce the movement of the predicted frame.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (9)
1. A method for detecting mycobacterium tuberculosis based on an improved YOLO v5 model, comprising:
inputting a tubercle bacillus image in a sputum sample into a trained improved YOLOv5 model, and detecting tubercle bacillus in the image;
the improved YOLOv5 model comprises: input end, improved Backbone network, FPNs network, head;
the training process for improving the YOLOv5 model comprises the following steps:
s1: acquiring a data set of a tubercle bacillus image in a sputum sample, marking the acquired data, performing format conversion treatment, and dividing a training set and the data set from the data set after format conversion;
s2: inputting the data in the training set to an input end of an improved YOLOv5 model for preprocessing;
s3: inputting the preprocessed picture into a Backbone network of a backhaul to obtain deep characteristic information, and obtaining characteristic diagrams with different depths;
s4: inputting feature maps with different depths into an FPNs network to perform up-sampling and feature fusion to obtain tensor data with different scales;
s5: tensor data of different scales are used for respectively predicting classification and prediction frames of targets through Head heads;
s6: selecting a correct detection frame according to the prediction frame and the real frame, and eliminating redundant prediction frames by using a non-maximum suppression NMS method;
s7: and calculating a loss function of the model according to the detection result, then adjusting parameters of the model, and finishing model training when the change of the loss function value is floating or not smaller.
2. The method for detecting tubercle bacillus based on the modified YOLO v5 model according to claim 1, wherein the modified Backbone network comprises: a Focus module and a CSPDarknet53 network;
merging a MHSA multi-head self-attention mechanism into each BottleNeck module of a C3 module of the last layer of the CSPDarknet53 network to obtain three BottleNeck modules of the last layer of the CSPDarknet53 network to become three BottleNeck Transformer modules;
the picture is divided into pixels through a Focus module of an improved CSPDarknet53 network, four matrixes which are unchanged in channel number and scaled into one fourth of the original size are obtained, the four input matrixes are then combined according to the channel number and input into the CSPDarknet53 network, and a characteristic diagram with the depth of 160 multiplied by 160, 80 multiplied by 80 and 40 multiplied by 40 is output through multi-layer convolution and MHSA multi-head self-attention mechanism operation.
3. The method for detecting tubercle bacillus based on the improved YOLO v5 model according to claim 1, wherein the FPNs network comprises:
target fusion detection was performed using a 160×160 large-scale detection layer, an 80×80 medium-scale detection layer, and a 40×40 small-scale detection layer.
4. The method for detecting tubercle bacillus based on the modified YOLO v5 model according to claim 1, wherein the preprocessing of the data in the training set by the input terminal comprises:
and carrying out random horizontal overturning, color adjustment, multi-scale input, sample mixing in proportion, cutting and splicing on the input picture to obtain the image with enhanced data.
5. The method for detecting tubercle bacillus based on the improved YOLO v5 model according to claim 1, wherein the classification of tensor data at different scales by Head prediction targets comprises:
conv operation is carried out on tensor data with different scales through convolution kernels with the channel number of 6 and 3 multiplied by 3 to obtain a confidence coefficient parameter of the target, a confidence coefficient threshold is set, when the obtained confidence coefficient parameter is larger than the confidence coefficient threshold, the detection target is tubercle bacillus, and otherwise, the detection target is abandoned.
6. The method for detecting tubercle bacillus based on the improved YOLO v5 model according to claim 1, wherein the prediction frame for predicting the target by Head on the feature map of different scales comprises:
conv operation is carried out on tensor data with different scales through convolution kernels with the channel number of 6 and 3 multiplied by 3, 4 position information parameters are obtained, and the parameters respectively represent the length and the width of a prediction frame and the vertical and horizontal distances of the prediction frame from the upper left corner of an image.
7. The method for detecting tubercle bacillus based on the improved YOLO v5 model according to claim 1, wherein selecting a correct detection frame based on a prediction frame and a real frame comprises:
and calculating IoU values of a prediction frame and a real frame generated by the model on the tubercle bacillus, taking the IoU value as a threshold value of whether the model identifies the tubercle bacillus, and when the prediction frame exceeds the IoU threshold value, considering that the correct tubercle bacillus is detected, so as to obtain a detection frame with the correct prediction frame.
9. The method for detecting tubercle bacillus based on the modified YOLO v5 model according to claim 1, wherein calculating the loss function of the model comprises:
wherein L is box A SIoU loss function representing the model, ioU represents IoU values of the model for the prediction box and the true box generated by the tubercle bacillus,b represents a prediction frame, B GT Representing the true box, delta represents the angular cost,ρ t indicating angle loss->C w Representing the length of a rectangle with the line connecting the center points of the real frame and the predicted frame as diagonal lines, C h A width of a rectangle representing that a line connecting a center point of a real frame and a predicted frame is a diagonal line, γ represents an angle loss offset, γ=2- Λ, Λ represents a distance cost, +.>Omega represents shape cost, < >>ω t Representing shape loss, ++>w、h、w gt 、h gt Representing the width and height of the predicted and real frames, respectively, θ represents the degree of attention to control the shape loss. />
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310222513.6A CN116229236A (en) | 2023-03-09 | 2023-03-09 | Bacillus tuberculosis detection method based on improved YOLO v5 model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310222513.6A CN116229236A (en) | 2023-03-09 | 2023-03-09 | Bacillus tuberculosis detection method based on improved YOLO v5 model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116229236A true CN116229236A (en) | 2023-06-06 |
Family
ID=86576639
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310222513.6A Pending CN116229236A (en) | 2023-03-09 | 2023-03-09 | Bacillus tuberculosis detection method based on improved YOLO v5 model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116229236A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116630720A (en) * | 2023-06-09 | 2023-08-22 | 中国电子技术标准化研究院 | Antigen detection result recognition method based on YoloV7 |
-
2023
- 2023-03-09 CN CN202310222513.6A patent/CN116229236A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116630720A (en) * | 2023-06-09 | 2023-08-22 | 中国电子技术标准化研究院 | Antigen detection result recognition method based on YoloV7 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111325739B (en) | Method and device for detecting lung focus and training method of image detection model | |
WO2019238063A1 (en) | Text detection and analysis method and apparatus, and device | |
WO2018108129A1 (en) | Method and apparatus for use in identifying object type, and electronic device | |
CN113592845A (en) | Defect detection method and device for battery coating and storage medium | |
CN111047609B (en) | Pneumonia focus segmentation method and device | |
WO2020024127A1 (en) | Bone age assessment and height prediction model, system thereof and prediction method therefor | |
CN110969245B (en) | Target detection model training method and device for medical image | |
WO2022012110A1 (en) | Method and system for recognizing cells in embryo light microscope image, and device and storage medium | |
CN114241548A (en) | Small target detection algorithm based on improved YOLOv5 | |
JP7391504B2 (en) | Information processing device, information processing method and program | |
WO2022148109A1 (en) | Product defect detection method and apparatus, device and computer-readable storage medium | |
WO2020077961A1 (en) | Image-based breast lesion identification method and device | |
CN116664559B (en) | Machine vision-based memory bank damage rapid detection method | |
CN112907519A (en) | Metal curved surface defect analysis system and method based on deep learning | |
US20210390282A1 (en) | Training data increment method, electronic apparatus and computer-readable medium | |
CN112634257B (en) | Fungus fluorescence detection method | |
US20240005494A1 (en) | Methods and systems for image quality assessment | |
US20090232400A1 (en) | Image evaluation apparatus, method, and program | |
CN116229236A (en) | Bacillus tuberculosis detection method based on improved YOLO v5 model | |
CN115359035A (en) | Lung X-ray focus detection system and method based on deformable multi-scale convolution | |
CN113782184A (en) | Cerebral apoplexy auxiliary evaluation system based on facial key point and feature pre-learning | |
CN113793385A (en) | Method and device for positioning fish head and fish tail | |
CN108416304B (en) | Three-classification face detection method using context information | |
CN111612749A (en) | Lung image-based focus detection method and device | |
CN113469942B (en) | CT image lesion detection method |
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 |