CN115063680A - Bridge disease identification method based on label and image synthesis technology - Google Patents
Bridge disease identification method based on label and image synthesis technology Download PDFInfo
- Publication number
- CN115063680A CN115063680A CN202210698546.3A CN202210698546A CN115063680A CN 115063680 A CN115063680 A CN 115063680A CN 202210698546 A CN202210698546 A CN 202210698546A CN 115063680 A CN115063680 A CN 115063680A
- Authority
- CN
- China
- Prior art keywords
- image
- disease
- bridge
- label
- foreground
- 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
- 201000010099 disease Diseases 0.000 title claims abstract description 124
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 124
- 238000000034 method Methods 0.000 title claims abstract description 28
- 230000015572 biosynthetic process Effects 0.000 title claims abstract description 26
- 238000003786 synthesis reaction Methods 0.000 title claims abstract description 26
- 238000005516 engineering process Methods 0.000 title claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 14
- 230000011218 segmentation Effects 0.000 claims abstract description 6
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 230000002194 synthesizing effect Effects 0.000 abstract description 4
- 238000013135 deep learning Methods 0.000 abstract description 2
- 238000004364 calculation method Methods 0.000 abstract 2
- 238000001514 detection method Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 4
- 238000002372 labelling Methods 0.000 description 4
- 238000003709 image segmentation Methods 0.000 description 3
- 238000013136 deep learning model Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000009440 infrastructure construction Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- 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
-
- 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
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- 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
- G06V10/20—Image preprocessing
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- 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
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20132—Image cropping
-
- 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
- G06T2207/20212—Image combination
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Databases & Information Systems (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a bridge disease identification method based on a label and an image synthesis technology, and belongs to the technical field of calculation, calculation or counting. The method comprises the following steps: acquiring images of a small number of bridge diseases, marking semantic information of the diseases by using an image marking tool, and acquiring disease image labels; performing mask operation on the disease image and the disease image label to obtain a foreground image of the disease; acquiring a bridge disease-free image; synthesizing a large number of bridge disease images by using a small number of disease foreground images and disease-free images after mask operation, and generating label data of the images by recording image synthesis positions; and training a disease segmentation network based on deep learning by using the synthesized image data and the label information. The invention synthesizes a large amount of bridge disease images with pixel-level marking by using a small amount of marking data and some disease-free data, reduces the cost of manual marking data, and improves the accuracy and efficiency of disease identification.
Description
Technical Field
The invention relates to the crossing field of bridge detection and image processing, and particularly discloses a bridge disease identification method based on a label and an image synthesis technology.
Background
In recent years, with the rapid development of infrastructure construction in China, a large number of roads and bridges are constructed, and the later detection and maintenance work of the bridges follows. At present, the main approach of bridge detection is manual detection, and the method is time-consuming and labor-consuming and can influence traffic. Some bridges built in mountains and on the sea are difficult to realize manual detection and to ensure the safety of bridge detection personnel. Therefore, an efficient and accurate bridge detection method is urgently needed.
With the rapid development of image processing technologies based on deep learning, in particular, image segmentation technologies based on supervised learning, are widely applied to disease detection of bridges, tunnels and roads in civil engineering. The method comprises the steps that a large number of labeled bridge disease images are needed to construct a data training set for identifying bridge diseases by utilizing a deep learning model, in the prior art, most of training data are labeled pixel by picture labeling tools such as labelme software to obtain image semantic segmentation labels, so that a large amount of manpower and material resources are consumed, and the timeliness of bridge disease identification cannot be met; on the other hand, enough rare bridge disease image data are not easily acquired in actual engineering, so that the number of samples of a bridge disease image data set is insufficient, the accuracy of deep learning model identification is affected, and the accuracy of bridge disease identification is further affected.
In summary, the present invention is directed to a bridge disease identification method based on a label synthesis technology, and aims to obtain a large number of bridge disease images with pixel-level labels through a simple digital synthesis technology.
Disclosure of Invention
The invention aims to provide a bridge disease identification method based on label and image synthesis technology, which aims to solve the technical problems that the existing label data training set acquisition mode cannot meet the requirements of accuracy and timeliness of bridge disease identification by realizing acquisition of a large number of synthetic images and corresponding pixel level label images through simple assignment operation.
The invention adopts the following technical scheme for realizing the aim of the invention:
a bridge disease identification method based on a label and an image synthesis technology comprises the following steps:
s1, acquiring images of a small number of bridge diseases, and marking semantic information of the diseases by using an image marking tool;
s2, performing mask operation on the original image collected in the step S1 and the generated label, keeping the pixels of the disease image area in the original image unchanged, and setting the pixels of the background part in the original image to zero to obtain a disease foreground image;
s3, converting the image obtained in the step S2 into an RGBA image;
s4, acquiring a bridge disease-free image, and converting the bridge disease-free image into an RGBA image;
s5, synthesizing a large number of images of the bridge diseases by using the images obtained in S3 and the images obtained in S4, and generating label data of the images by recording the image synthesis positions;
and S6, training an image segmentation network by using the synthetic bridge disease image and the label data generated in S5, and realizing quantitative identification of the bridge disease.
Preferably, in step S1, a high-definition camera is used to collect the bridge disease image, and it is ensured that the real physical size corresponding to each pixel is not less than 0.1 mm under the condition that the shooting distance is 50 cm.
Preferably, the image annotation tool used in step S1 can generate annotations at the bridge pixel level and ensure that the label and the image have the same resolution.
Preferably, the number of disease images of a certain type acquired and labeled in step S1 is generally no more than 10% of the number of final composite image data sets.
Preferably, after the bridge original image and the pixel-level label are masked in step S2, the size and resolution of the image remain unchanged.
Preferably, in step S3, the value of the transparency channel of the background portion of the image is adjusted to 0, and the transparency of the damaged portion is adjusted to 1.
Preferably, the resolution of the bridge disease-free image acquired in step S4 is not lower than that of the image acquired in step S1, the number of images is not lower than 200, and the images need to be transparent images.
Preferably, in step S5, the disease foreground image in the RGBA format is attached to the background image with all RGBA channels being zero to obtain a synthesized foreground image, the synthesized foreground image is added to the bridge disease-free image in the RGBA format to obtain a synthesized image, meanwhile, an assignment operation is performed on the synthesized foreground image to obtain a label image of the synthesized image, and the synthesized image and the label image form a data training set.
The specific method for obtaining the synthetic foreground image by pasting the disease foreground image in the RGBA format to the background image with the RGBA channels being zero comprises the following steps:
randomly cutting a disease foreground image in an RGBA format to obtain a foreground image;
randomly cutting the bridge disease-free image in the RGBA format to obtain a background image, generating an image which has the same size with the background image but all pixels are zero, and recording the image which has the same size with the background image but all pixels are zero as the background image with RGBA channels being zero.
Assigning the foreground image to a background image with RGBA channels being zero in a random lofting mode, and acquiring a synthetic foreground image
Preferably, in step S6, the trained neural network may be an image segmentation network, such as deplab V3 +.
By adopting the technical scheme, the invention has the following beneficial effects:
(1) according to the invention, on the premise of limited data volume by sea, a large number of bridge disease images and corresponding labels can be synthesized by a digital synthesis technology on the premise of a small number of labels, so that the cost of manual labeling data is greatly reduced, and the accuracy and efficiency of bridge disease identification are improved.
(2) The invention adopts a digital synthesis technology for synthesizing a large number of bridge disease images, the disease foreground images are pasted into the background images with zero RGBA channels in a random lofting mode to obtain the synthetic foreground images, the types of the synthetic foreground images can be enriched by randomly cutting the disease foreground images and the background images, the binary operation is carried out on the synthetic foreground images in the process of synthesizing the synthetic foreground images and the background images, the synthetic foreground images are converted into label images for recording the synthetic position information, further the synthetic images and the pixel level label images corresponding to the synthetic objects are obtained, a training data set for identifying bridge diseases is constructed by the synthetic images and the corresponding pixel level labels, sufficient training data can improve the identification accuracy, and the time for obtaining the training data set is greatly reduced compared with the traditional manual labeling mode, and further improve the efficiency of recognition.
Drawings
FIG. 1 is a flow chart of bridge defect identification according to the present invention.
Fig. 2(a) is a crack image, and fig. 2(b) is an image label obtained by labeling the crack image shown in fig. 2 (a).
FIG. 3 is a diagram illustrating a randomly cropped image according to the present invention.
FIG. 4 is a schematic diagram of an image synthesis technique according to the present invention.
Fig. 5 is a schematic diagram of a deep neural network used in the present invention.
Detailed Description
The technical scheme of the invention is explained in detail in the following with reference to the attached drawings.
As shown in fig. 1, the invention discloses a bridge disease identification method based on a label and an image synthesis technology. Including the following steps S1 to S6.
Step S1: acquiring a small number of bridge disease images, and marking semantic information of the disease images by using an image marking tool
Under the condition that the shooting distance is guaranteed to be 50cm, a high-definition camera is adopted to collect bridge disease images, and the real physical size corresponding to each pixel is not less than 0.1 mm; and selecting 50 images from the shot bridge disease images, and performing semantic annotation on each selected image by adopting a labelme tool to obtain the label of each bridge disease image. And (3) performing pixel-by-pixel semantic annotation on the crack image shown in the fig. 2(a) by using an annotation tool to obtain the image label shown in the fig. 2 (b).
Step S2: performing a mask operation on the original bridge disease image collected in step S1 and the image label generated in step S1
And for each original bridge disease image, performing mask operation on the original bridge disease image and the corresponding pixel level image label thereof, keeping the size and the resolution of the image unchanged, keeping the disease image area in the original bridge disease image unchanged, and setting pixels of the area except the disease image area in the original bridge disease image to zero to obtain a disease foreground image I.
Step S3: converting the image obtained after the mask operation in the step S2 into an RGBA image
Adjusting the numerical value of the background partial transparency channel of the image obtained after the mask operation of the step S2 to 0; and adjusting the partial transparency of the disease image area to 1 to obtain a disease foreground image I in an RGBA format.
Step S4: acquiring a bridge disease-free image, and converting the acquired bridge disease-free image into an RGBA image
The acquired bridge disease-free images are images with transparency and the resolution is not lower than that of the bridge disease images acquired in the step S1, and the number of the acquired bridge disease-free images is not lower than 200.
Step S5: using the image obtained in step S3 and the image obtained in step S4, images of bridge defects are synthesized in large quantities, and label data of the images is generated by recording the image synthesis positions
As shown in fig. 4, firstly, randomly cutting the disease foreground image I obtained in step S3 to obtain an image I0, randomly cutting the bridge disease-free image obtained in step S4 to obtain a background image Ib, then generating an image I1 which has the same size as the background image Ib but each pixel is zero, assigning the value of I0 obtained by randomly cutting the disease foreground image to I1 by means of random lofting to obtain a synthesized foreground image I2, generating a label image mask by the synthesized foreground image I2 through pixel-by-pixel operation, and specifically, the pixel-by-pixel operation of the synthesized foreground image I2 is to set only a non-zero pixel value to 1 while keeping zero pixels in the synthesized foreground image I2 unchanged; subsequently, the synthetic foreground image I2 is added to the background image Ib to obtain a synthetic image Ic, thereby obtaining a synthetic image and a corresponding label.
Fig. 3 shows a schematic diagram of randomly cropping an image, and an image generated by randomly cropping the image generated in step S3 is attached to an image generated by randomly cropping the image generated in step S4 and synthesized, so that the diversity of the synthesized foreground images is increased. The position of the image obtained in the step S3 attached to the image obtained in the step S4 after cutting is recorded in the synthetic foreground image I2 obtained through random lofting, and the image I2 is binarized to obtain a label image corresponding to the synthetic image. When a plurality of diseases are contained in the disease foreground image, the image I2 is assigned with a value according to the type of the contained disease, and a label image recording the image synthesis position (the position in the foreground image corresponding to each disease) can be obtained.
Step S6: training semantic segmentation network by using data training set generated by S5 to realize quantitative identification of bridge diseases
The semantic segmentation network may be selected as deep V3+ as shown in fig. 5.
The above is only a preferred embodiment of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and improvements can be made without departing from the spirit of the invention, and these modifications and improvements are intended to be within the scope of the invention.
Claims (10)
1. A bridge disease identification method based on label and image synthesis technology is characterized in that,
acquiring a bridge disease image, and performing semantic information annotation on the bridge disease image to acquire a label of the bridge disease image;
performing mask operation on the bridge disease image and the label thereof to obtain a disease foreground image and obtain a disease foreground image in an RGBA format;
acquiring a bridge disease-free image, and converting the acquired bridge disease-free image into an RGBA format;
attaching the disease foreground image in the RGBA format to a background image with zero RGBA channels to obtain a synthetic foreground image, adding the synthetic foreground image and the bridge disease-free image in the RGBA format to obtain a synthetic image, and performing assignment operation on the synthetic foreground image to obtain a label image of the synthetic image, wherein the synthetic image and the label image form a data training set;
and training a neural network for identifying the bridge diseases by adopting the data training set.
2. The bridge disease identification method based on the label and image synthesis technology according to claim 1, wherein the specific method for obtaining the synthesized foreground image by pasting the disease foreground image in the RGBA format to the background image with the RGBA channels being zero comprises the following steps:
randomly cutting the disease foreground image in the RGBA format to obtain a foreground image;
randomly cutting the bridge disease-free image in the RGBA format to obtain a background image, generating an image which has the same size as the background image and all pixels are zero, and recording the image which has the same size as the background image and all pixels are zero as the background image with RGBA channels being zero;
and assigning the foreground image to a background image with zero RGBA channels in a random lofting mode to obtain a synthetic foreground image.
3. The bridge disease identification method based on the label and image synthesis technology according to claim 1, wherein the specific method for obtaining the label image of the synthetic image by performing the assignment operation on the synthetic foreground image is as follows: and performing pixel-by-pixel operation on the synthesized foreground image, keeping 0 pixel in the synthesized foreground image unchanged, setting a non-zero pixel in the synthesized foreground image as a corresponding number according to the type of the disease, and acquiring a label image of the synthesized image.
4. The bridge disease identification method based on the label and image synthesis technology as claimed in claim 1, wherein a high-definition camera is adopted to collect bridge disease images under the condition that the shooting distance is guaranteed to be 50cm, and the real physical size corresponding to each pixel is greater than or equal to 0.1 mm.
5. The bridge disease identification method based on the label and image synthesis technology as claimed in claim 1, wherein the number of the acquired and labeled certain bridge disease images is less than or equal to 10% of the number of the final synthetic image data sets.
6. The bridge disease identification method based on the label and image synthesis technology according to claim 1, wherein the specific method for performing mask operation on the bridge disease image and the label thereof to obtain the disease foreground image is as follows: keeping the size and the resolution of the bridge disease image unchanged, keeping the pixel value of the disease image area in the bridge disease image unchanged, setting the pixel value of the background area in the bridge disease image to 0, and obtaining a disease foreground image.
7. The bridge disease identification method based on the label and image synthesis technology according to claim 1, wherein the specific method for obtaining the disease foreground image in the RGBA format is as follows: and adjusting the transparency of the background part of the disease foreground image to be 0, and adjusting the transparency of the disease image area in the disease foreground image to be 1.
8. The bridge disease identification method based on the label and image synthesis technology as claimed in claim 1, wherein the bridge disease-free image has transparency and resolution higher than or equal to the bridge disease image, and the number of the acquired bridge disease-free images is greater than or equal to 200.
9. The bridge disease identification method based on the label and image synthesis technology as claimed in claim 1, wherein the neural network for identifying the bridge disease is a semantic segmentation network.
10. The bridge disease identification method based on the label and image synthesis technology according to claim 9, wherein the semantic segmentation network is deep V3 +.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210698546.3A CN115063680B (en) | 2022-06-20 | 2022-06-20 | Bridge disease identification method based on label and image synthesis technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210698546.3A CN115063680B (en) | 2022-06-20 | 2022-06-20 | Bridge disease identification method based on label and image synthesis technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115063680A true CN115063680A (en) | 2022-09-16 |
| CN115063680B CN115063680B (en) | 2024-05-17 |
Family
ID=83202582
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210698546.3A Active CN115063680B (en) | 2022-06-20 | 2022-06-20 | Bridge disease identification method based on label and image synthesis technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115063680B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116168242A (en) * | 2023-02-08 | 2023-05-26 | 阿里巴巴（中国）有限公司 | Pixel-level label generation method, model training method and equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150294477A1 (en) * | 2014-04-09 | 2015-10-15 | Google Inc. | Image-Based Bridge Identification and Boundary Detection |
| CN111861978A (en) * | 2020-05-29 | 2020-10-30 | 陕西师范大学 | Bridge crack example segmentation method based on Faster R-CNN |
| CN112488990A (en) * | 2020-11-02 | 2021-03-12 | 东南大学 | Bridge bearing fault identification method based on attention regularization mechanism |
-
2022
- 2022-06-20 CN CN202210698546.3A patent/CN115063680B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150294477A1 (en) * | 2014-04-09 | 2015-10-15 | Google Inc. | Image-Based Bridge Identification and Boundary Detection |
| CN111861978A (en) * | 2020-05-29 | 2020-10-30 | 陕西师范大学 | Bridge crack example segmentation method based on Faster R-CNN |
| CN112488990A (en) * | 2020-11-02 | 2021-03-12 | 东南大学 | Bridge bearing fault identification method based on attention regularization mechanism |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116168242A (en) * | 2023-02-08 | 2023-05-26 | 阿里巴巴（中国）有限公司 | Pixel-level label generation method, model training method and equipment |
| CN116168242B (en) * | 2023-02-08 | 2023-12-01 | 阿里巴巴（中国）有限公司 | Pixel-level label generation method, model training method and equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115063680B (en) | 2024-05-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111967313B (en) | Unmanned aerial vehicle image annotation method assisted by deep learning target detection algorithm | |
| CN111899288A (en) | Tunnel leakage water area detection and identification method based on infrared and visible light image fusion | |
| CN113674216A (en) | Subway tunnel disease detection method based on deep learning | |
| CN115063680B (en) | Bridge disease identification method based on label and image synthesis technology | |
| CN110363769B (en) | Image segmentation method for cantilever system of high-speed rail contact net supporting device | |
| CN117291902B (en) | Detection method for pixel-level concrete cracks based on deep learning | |
| CN112102250B (en) | Method for establishing and detecting pathological image detection model with training data as missing label | |
| CN110378916A (en) | A kind of TBM image based on multitask deep learning is slagged tap dividing method | |
| CN116597270A (en) | Road damage target detection method based on attention mechanism integrated learning network | |
| CN107886125A (en) | MODIS satellite remote sensing images mask methods based on local spectral factorization marking | |
| CN113780117B (en) | Method for rapidly identifying and extracting relevant parameters of estuary plume outline | |
| CN118196028A (en) | Rural cement pavement disease extraction method with improved YOLOv8 | |
| CN117710843A (en) | Intersection dynamic signal timing scheme detection method based on unmanned aerial vehicle video | |
| CN111626971B (en) | Smart city CIM real-time imaging method with image semantic perception | |
| CN117078925A (en) | Building rubbish annual output accurate calculation method based on RDSA-deep LabV3+ network | |
| CN110956174A (en) | Device number identification method | |
| CN111047646A (en) | Multi-target lens positioning method and system based on FPGA | |
| CN114494862B (en) | Regional fast-growing forest density accurate statistical method based on unmanned aerial vehicle image | |
| CN110363198A (en) | A kind of neural network weight matrix fractionation and combined method | |
| CN112966774B (en) | Picture Bert-based tissue pathology picture classification method | |
| CN115797904A (en) | Active learning method for multiple scenes and multiple tasks in intelligent driving visual perception | |
| CN114429573A (en) | Data enhancement-based household garbage data set generation method | |
| CN115249319A (en) | Method for detecting sun dark stripes in full-sun-surface image | |
| CN114529879B (en) | Mine electric locomotive road condition real-time detection method based on YOLOv-Tiny | |
| CN110032997B (en) | Natural scene text positioning method based on image segmentation |
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 |