CN112529830B - Image labeling method, device, electronic equipment and storage medium - Google Patents
Image labeling method, device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN112529830B CN112529830B CN201910814719.1A CN201910814719A CN112529830B CN 112529830 B CN112529830 B CN 112529830B CN 201910814719 A CN201910814719 A CN 201910814719A CN 112529830 B CN112529830 B CN 112529830B
- Authority
- CN
- China
- Prior art keywords
- hole
- foreground region
- image
- binary image
- marked
- 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.)
- Active
Links
- 238000002372 labelling Methods 0.000 title claims abstract description 90
- 238000000034 method Methods 0.000 claims abstract description 30
- 238000012545 processing Methods 0.000 claims description 33
- 238000005260 corrosion Methods 0.000 claims description 19
- 230000007797 corrosion Effects 0.000 claims description 19
- 238000010586 diagram Methods 0.000 description 12
- 238000012549 training Methods 0.000 description 6
- 238000005530 etching Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 230000003628 erosive effect Effects 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The application provides an image labeling method, an image labeling device, electronic equipment and a storage medium, wherein the method comprises the following steps: generating a foreground region corresponding to the target object according to the pixel coordinates of the target object in the marked image; generating a hypothesis hole in the foreground region, and updating the foreground region according to a judging result of whether the hypothesis hole is a real hole, wherein the updated foreground region comprises the real hole; and marking the pixel blocks corresponding to the pixel coordinates as holes in the marked image according to the pixel coordinates of the real holes in the updated foreground region. The image labeling method provided by the application can determine the true cavity in the foreground area labeled as the target object, so that the cavity in the target object can be labeled again, and the accuracy of image labeling is improved.
Description
Technical Field
The present application relates to the field of image labeling technologies, and in particular, to an image labeling method, an image labeling device, an electronic device, and a storage medium.
Background
Image annotation is a process of classifying blocks of pixels contained in an image, which is essentially classifying blocks of pixels belonging to different objects in the image. For example, it is noted which pixel blocks belong to a person, which pixel blocks belong to a tree, which pixel blocks belong to a lane, and so on. At present, an image annotation model is mostly adopted to annotate an image. The image annotation model is obtained by training a training data set, wherein the training data set comprises a large number of annotated images.
In the prior art, in order to ensure the accuracy of an image annotation model obtained by training, an image is often annotated in a manual annotation mode so as to obtain an accurate training data set. The same object in the image may contain a hole, and the corresponding pixel block in the hole belongs to other objects. For example, the image may include voids between the branches and leaves of the tree, which may be sky, house, etc., rather than the tree. In order to save labor, all pixel blocks containing holes are marked as the same object during manual marking, for example, all pixel blocks containing holes are marked as trees. However, the labeling accuracy of the labeling mode is low, and the training result of the image labeling model can be adversely affected.
Disclosure of Invention
The application provides an image labeling method, an image labeling device, electronic equipment and a storage medium, which can determine real holes in a foreground area labeled as a target object, so that the holes in the target object can be labeled again, and the accuracy of image labeling is improved.
A first aspect of the present application provides an image labeling method, including:
generating a foreground region corresponding to a target object in the marked image according to the pixel coordinates of the target object;
Generating a hypothesis hole in the foreground region, updating the foreground region according to a judging result of whether the hypothesis hole is a real hole, wherein the updated foreground region comprises the real hole;
and marking the pixel blocks corresponding to the pixel coordinates as holes in the marked image according to the pixel coordinates of the real holes in the updated foreground region.
The foreground region is a foreground region in a binary image, the generating a hypothetical cavity in the foreground region, and updating the foreground region according to a determination result of whether the hypothetical cavity is a real cavity, including:
generating a hypothesis hole in a foreground region in the binary image, and updating the binary image according to a judging result of whether the hypothesis hole is a real hole, wherein the foreground region in the updated binary image is the updated foreground region;
and marking the pixel block corresponding to the pixel coordinates as a hole in the marked image according to the pixel coordinates of the real hole in the updated foreground region, wherein the marking comprises the following steps:
and marking the pixel block corresponding to the pixel coordinates as a hole in the marked image according to the pixel coordinates of the real hole in the updated binary image.
Optionally, the generating a hypothesis hole in the foreground area in the binary image, and updating the binary image according to a determination result of whether the hypothesis hole is a true hole, includes:
A. binary image P corresponding to iteration period i i Randomly generating a hypothesis cavity according to a preset probability, wherein i is an integer greater than or equal to 1;
B. judging the P in the marked image according to the pixel coordinates of the assumed holes and the corresponding relation between the distribution of the assumed holes and the distribution of the real holes i Whether the assumed hole on the foreground region is a real hole;
C. according to the P i Judging whether the assumed cavity on the foreground region is a true cavity or not according to the judgment result of the true cavity, and judging the P i Marking, corroding and expanding in sequence to obtain an updated binary image P corresponding to the iteration period i i ”；
D. Judging whether the i is equal to a preset value, if so, executing E, and if not, executing the P i "as the binary image corresponding to the next iteration period, and executing the A in a return manner;
E. the updated binary image P corresponding to the iteration period i is processed i "as the updated binary image.
Optionally, the method comprises the following steps of i Judging whether the assumed cavity on the foreground region is a true cavity or not according to the judgment result of the true cavity, and judging the P i Marking, corroding and expanding in sequence to obtain an updated binary image P corresponding to the iteration period i i ", comprising:
the P is set i The pixel block corresponding to the assumed hole which is the real hole on the foreground area is marked as the background area, and the P is marked i The pixel blocks corresponding to the assumed holes which are not real holes on the foreground region are marked as the foreground region, and a marked binary image corresponding to the iteration period i is obtained;
performing corrosion treatment on the marked binary image corresponding to the iteration period i to obtain a corrosion treated binary image P corresponding to the iteration period i i '；
Corresponding to the iteration period iBinary pattern P after etching treatment i ' performing expansion treatment to obtain the P i ”。
Optionally, before labeling the pixel block corresponding to the pixel coordinate as the hole in the labeled image according to the pixel coordinate of the real hole in the updated binary image, the method further includes:
obtaining pixel coordinates of a real hole in the updated binary image by comparing a foreground area corresponding to a first iteration period with a foreground area corresponding to a last iteration period, wherein the pixel coordinates of the real hole are as follows: in the first iteration period, the pixel coordinates corresponding to the pixel blocks belonging to the foreground region in the last iteration period.
Before the generating of the hypothesis hole in the foreground region in the binary image, the method further comprises:
receiving a first instruction input by a user, wherein the first instruction is used for indicating to expand a foreground region of the binary image, and is input when the labeling contour area of a target object in the labeled image is smaller than a first preset contour area; and/or the number of the groups of groups,
and receiving a second instruction input by the user, wherein the second instruction is used for indicating the foreground region of the binary image to be corroded, and the second instruction is input when the marked contour area of the target object in the marked image is larger than a second preset contour area, and the second preset contour area is larger than the first preset contour area.
Optionally, after labeling the pixel block corresponding to the pixel coordinate as the hole in the labeled image according to the pixel coordinate of the real hole in the updated binary image, the method further includes:
receiving an annotation instruction input by a user, wherein the annotation instruction comprises the following steps: an object corresponding to a target pixel block in the pixel blocks marked as holes in the marked image;
and marking the target pixel block as an object corresponding to the target pixel block according to the marking instruction.
A second aspect of the present application provides an image labeling apparatus comprising:
the first processing module is used for generating a foreground area corresponding to the target object according to the pixel coordinates of the target object in the marked image; generating a hypothesis hole in the foreground region, updating the foreground region according to a judging result of whether the hypothesis hole is a real hole, wherein the updated foreground region comprises the real hole;
and the second processing module is used for marking the pixel block corresponding to the pixel coordinates as a hole in the marked image according to the pixel coordinates of the real hole in the updated foreground region.
Optionally, the foreground region is a foreground region in a binary image.
Correspondingly, the first processing module is specifically configured to generate a hypothetical hole in a foreground area in the binary image, update the binary image according to a determination result of whether the hypothetical hole is a real hole, and the foreground area in the updated binary image is the updated foreground area.
And the second processing module is used for marking the pixel block corresponding to the pixel coordinate as a hole in the marked image according to the pixel coordinate of the real hole in the updated binary image.
Optionally, the first processing module is specifically configured to:
A. binary image P corresponding to iteration period i i Randomly generating a hypothesis cavity according to a preset probability, wherein i is an integer greater than or equal to 1;
B. judging the P in the marked image according to the pixel coordinates of the assumed holes and the corresponding relation between the distribution of the assumed holes and the distribution of the real holes i Whether the assumed hole on the foreground region is a real hole;
C. according to the P i Judging whether the assumed cavity on the foreground region is a true cavity or not according to the judgment result of the true cavity, and judging the P i Marking, corroding and expanding in sequence to obtain an updated binary image P corresponding to the iteration period i i ”；
D. Judging whether the i is equal to a preset value, if so, executing E, and if not, executing the P i "as the binary image corresponding to the next iteration period, and executing the A in a return manner;
E. the updated binary image P corresponding to the iteration period i is processed i "as the updated binary image.
Optionally, the first processing module is specifically configured to: the P is set i The pixel block corresponding to the assumed hole which is the real hole on the foreground area is marked as the background area, and the P is marked i The pixel blocks corresponding to the assumed holes which are not real holes on the foreground region are marked as the foreground region, and a marked binary image corresponding to the iteration period i is obtained; performing corrosion treatment on the marked binary image corresponding to the iteration period i to obtain a corrosion treated binary image P corresponding to the iteration period i i 'A'; a binary image P corresponding to the iteration period i after corrosion treatment i ' performing expansion treatment to obtain the P i ”。
Optionally, the first processing module is further configured to obtain pixel coordinates of a real hole in the updated binary image by comparing a foreground area corresponding to a first iteration period with a foreground area corresponding to a last iteration period, where the pixel coordinates of the real hole are: in the first iteration period, the pixel coordinates corresponding to the pixel blocks belonging to the foreground region in the last iteration period.
Optionally, the image labeling device further includes: and a transceiver module.
The receiving and transmitting module is used for receiving a first instruction input by a user before the foreground region generates a hypothesis hole, wherein the first instruction is used for indicating to expand the foreground region of the binary image, and the first instruction is input when the labeling contour area of a target object in the labeled image is smaller than a first preset contour area; and/or the number of the groups of groups,
The second instruction is used for indicating the foreground area of the binary image to be corroded, and is input when the marked contour area of the target object in the marked image is larger than a second preset contour area, and the second preset contour area is larger than the first preset contour area.
Optionally, the transceiver module is further configured to receive an annotation instruction input by a user, where the annotation instruction includes: and the object corresponding to the target pixel block in the pixel blocks marked as the holes in the marked image.
Correspondingly, the second processing module is configured to label the target pixel block as an object corresponding to the target pixel block according to the labeling instruction.
A third aspect of the present application provides an electronic apparatus comprising: at least one processor and memory;
the memory stores computer-executable instructions;
the at least one processor executes the computer-executable instructions stored by the memory to cause the electronic device to perform the image annotation method described above.
A fourth aspect of the present application provides a computer-readable storage medium having stored thereon computer-executable instructions that, when executed by a processor, implement the above-described image labeling method.
The application provides an image labeling method, an image labeling device, electronic equipment and a storage medium, wherein the method comprises the following steps: generating a foreground region corresponding to the target object according to the pixel coordinates of the target object in the marked image; generating a hypothesis hole in the foreground region, and updating the foreground region according to a judging result of whether the hypothesis hole is a real hole, wherein the updated foreground region comprises the real hole; and marking the pixel blocks corresponding to the pixel coordinates as holes in the marked image according to the pixel coordinates of the real holes in the updated foreground region. The image labeling method provided by the application can determine the true cavity in the foreground area labeled as the target object, so that the cavity in the target object can be labeled again, and the accuracy of image labeling is improved.
Drawings
FIG. 1 is a schematic flow chart of an image labeling method according to the present application;
FIG. 2 is a schematic diagram of an unlabeled image provided by the present application;
FIG. 3 is a schematic diagram of a binary image corresponding to a labeled image according to the present application;
FIG. 4 is a schematic diagram of an updated binary image according to the present application;
FIG. 5 is a second flow chart of the image labeling method according to the present application;
FIG. 6 is a schematic diagram of an image labeling apparatus according to the present application;
fig. 7 is a schematic structural diagram of an electronic device provided by the present application.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present application more apparent, the technical solutions in the embodiments of the present application will be clearly and completely described in the following in conjunction with the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
In order to solve the problem of low image labeling accuracy in the prior art, the application provides an image labeling method, which is used for labeling the cavities in the target object in the labeled image by determining the real cavities of the foreground area corresponding to the target object in the labeled image and further re-labeling the labeled image according to the pixel coordinates of the real cavities so as to achieve the aim of improving the labeling accuracy.
Fig. 1 is a schematic flow chart of an image labeling method provided by the present application. The main implementation of the method flow shown in fig. 1 may be an image labeling device, which may be implemented by any software and/or hardware. As shown in fig. 1, the image labeling method provided in this embodiment may include:
S101, generating a foreground region corresponding to the target object according to the pixel coordinates of the target object in the marked image.
The marked image in this embodiment is a coarsely marked image. Wherein, the rough labeling refers to labeling according to the labeling technology in the prior art, and optionally, the labeling technology can be manual labeling or labeling according to an image labeling model. And the noted image refers to a cavity in the unlabeled target object, and the target object and the cavity contained therein are both noted as target objects. The labeling method of the labeled image in this embodiment is not limited.
The target object is a preset object, and the preset object is an object which usually carries a cavity. Such as trees, fences, bicycles, etc. In this embodiment, after the image containing the target object is coarsely labeled, the pixel coordinates of the target object in the labeled image may be determined. For example, if the target object is a tree, then marking the tree in the image is: the pixel blocks belonging to the tree are labeled as trees. Correspondingly, the pixel blocks belonging to the tree can be determined in the marked image, and the pixel coordinates of the tree can be obtained.
In this embodiment, the foreground region of the target object may be generated according to the pixel coordinates of the target object in the annotated image. The foreground region of the target object is a region formed by pixel blocks corresponding to the target object in the marked image. In this embodiment, in order to facilitate the correspondence between the foreground region and the pixel block in the labeled image, the ratio of the pixel block may be 1:1 generates a foreground region of the target object.
In one possible implementation manner, the foreground area is a foreground area in a binary image, that is, in this embodiment, a binary image corresponding to the labeled image is generated, where the foreground area of the binary image is a pixel block corresponding to the target object. In order to facilitate the correspondence between the binary image and the pixel blocks in the marked image, the binary image of the birthday may have the same size as the marked image.
It should be understood that the two-value map is a background region except for the pixel block region (foreground region) corresponding to the target object. Alternatively, the gray value of the foreground area in this embodiment may be 0 or 255, that is, the foreground area may be black or white, and the corresponding gray value of the background area may be 255 or 0, that is, the background area may be white or black.
Illustratively, according to the pixel coordinates of the tree in the marked image, a binary image with the same size as the marked image is generated, and the foreground area of the binary image is an area formed by pixel blocks corresponding to the pixel coordinates of the tree. Fig. 2 is a schematic diagram of an unlabeled image provided by the present application, and fig. 3 is a schematic diagram of a binary image corresponding to the labeled image provided by the present application. Fig. 2 is an unlabeled image, and a binary image corresponding to the image in fig. 2 after being subjected to rough labeling is shown in fig. 3. As shown in fig. 3, the target object is a tree, and all the holes possibly carried in the tree are labeled as trees in the prior art (i.e. the tree in fig. 3 is not labeled with holes). It should be appreciated that the foreground region in the generated binary image is illustratively shown in fig. 3 as black.
Alternatively, in this embodiment, the binary image corresponding to the marked image (the rough marked image) may be marked finely.
By way of example, rough labeling of the target object in the image can determine the labeling contour area of the target object in the labeled image, and if the rough labeling result is accurate, the labeling contour area of the target object in the labeled image should be within the preset contour area range. The predetermined profile area range may be greater than the first predetermined profile area and less than a second predetermined profile area, wherein the second predetermined profile area is greater than the first predetermined profile area.
Optionally, in this embodiment, a first instruction input by the user may be received, where the first instruction is used to instruct to perform expansion processing on a foreground area of the binary image. It is noted that the first instruction is input when the labeling contour area of the target object in the labeled image is smaller than the first preset contour area.
If the user determines that the area of the foreground region in the binary image, namely the labeling outline area of the target object is smaller than the first preset outline area, because the rough labeling does not label the target object in the image completely, the foreground region of the binary image is subjected to expansion processing, so that the foreground region corresponding to the target object is enlarged, and the accuracy of the rough labeling result is improved. If the user determines that the labeling contour area of the target object is smaller than the first preset contour area, a first instruction can be input to instruct expansion processing of the foreground area of the binary image. Optionally, the first instruction includes an expansion factor, and the image labeling device may perform corresponding expansion processing on the foreground area of the binary image according to the expansion factor.
Optionally, in this embodiment, a second instruction input by the user may be received, where the second instruction is used to instruct corrosion processing on the foreground area of the binary image, and the second instruction is input when the labeled contour area of the target object in the labeled image is greater than the second preset contour area.
For example, if the user determines that the area of the foreground region in the binary image, that is, the labeling contour area of the target object is greater than the second preset contour area, since the rough labeling labels a plurality of users around the target object in the image and pixel blocks of the target object as the target object, the foreground region of the binary image needs to be corroded, so as to reduce the foreground region corresponding to the target object and improve the accuracy of the rough labeling result. If the user determines that the labeling contour area of the target object is larger than the second preset contour area, a second instruction can be input to indicate that the foreground area of the binary image is corroded. Optionally, the second instruction includes a corrosion factor, and the image labeling device may perform corresponding corrosion treatment on the foreground area of the binary image according to the corrosion factor. It should be appreciated that the first instruction and the second instruction described above may be received simultaneously when the target object is at least one.
S102, generating a hypothesis hole in the foreground region, and updating the foreground region according to a judging result of whether the hypothesis hole is a real hole, wherein the updated foreground region comprises the real hole.
In this embodiment, a hypothesis hole may be generated in the foreground region, where the hypothesis hole is randomly generated. Alternatively, in this embodiment, a hypothesis hole may be generated in the foreground region using a preset probability. For example, if the preset probability is 2%, the corresponding hypothesis hole may be generated with a probability of 2% in the pixel block corresponding to the foreground region. If there are 100 pixel blocks corresponding to the foreground region, 2 pixel blocks are determined as pixel blocks corresponding to the hole in the 100 pixel blocks.
After generating the hypothesis hole in the foreground region, it is necessary to determine whether the generated hypothesis hole is a true hole. In this embodiment, a conditional random field (Conditional random field algorithm, CRF) algorithm may be used to determine whether the hypothesized hole is a real hole. The judging result of whether the assumed hole is a real hole may include: which hypothetical holes are true holes and which hypothetical holes are not true holes.
It should be understood that, in the CRF algorithm adopted in this embodiment, on the premise that the pixel coordinates of the assumed holes are known, that is, after the distribution situation of the assumed holes is obtained, the distribution of the real holes can be predicted, that is, the pixel coordinates of the real holes can be determined, so that whether the assumed holes are real holes can be determined, and further, the judgment result of whether the assumed holes are real holes can be obtained. The principle of the CRF algorithm is not described in detail in this embodiment.
In this embodiment, the foreground region may be updated according to the determination result of whether the assumed hole is a real hole, to obtain an updated foreground region. The updating of the foreground region may be to mark a pixel block corresponding to an assumed hole determined as a true hole in the foreground region as a hole. It is noted that the updated foreground region in this embodiment includes a real hole.
In a possible implementation manner, when the foreground region is the foreground region in the binary image, the "generating a hypothetical hole in the foreground region and updating the foreground region according to the determination result of whether the hypothetical hole is a real hole" in the above embodiment is that "generating a hypothetical hole in the foreground region in the binary image and updating the binary image according to the determination result of whether the hypothetical hole is a real hole", where the foreground region in the updated binary image is the updated foreground region ". Correspondingly, the above-mentioned "the pixel block corresponding to the pixel coordinate is marked as a hole in the marked image according to the pixel coordinate of the true hole in the updated foreground region", that is, "the pixel block corresponding to the pixel coordinate is marked as a hole in the marked image according to the pixel coordinate of the true hole in the updated binary image".
The updated binary diagram corresponding to fig. 3 is shown in fig. 4, and fig. 4 is a schematic diagram of the updated binary diagram provided by the present application. In this embodiment, real holes in the tree in the foreground region can be marked in the updated binary image, and the holes are shown as white parts in the black tree in fig. 4.
And S103, marking the pixel block corresponding to the pixel coordinate as a hole in the marked image according to the pixel coordinate of the real hole in the updated foreground region.
In this embodiment, after the updated foreground region is obtained, the pixel coordinates of the real holes in the foreground region may be determined. According to the pixel coordinates of the real holes in the updated foreground region, the pixel blocks corresponding to the pixel coordinates can be marked as holes in the marked image.
It should be understood that, since the pixel blocks in the updated foreground region have an object relationship with the pixel blocks in the labeled image, the pixel block corresponding to the pixel coordinates of the real hole may be determined as the real hole in the labeled image, and then the pixel block corresponding to the pixel coordinates may be re-labeled as the hole.
For example, if the pixel block a in the updated foreground region is a real hole, the pixel block corresponding to the pixel block a in the marked image is marked as a real hole. The pixel block corresponding to the pixel block a is the pixel block with the same pixel coordinates.
In one possible implementation manner, when the foreground region is a foreground region in the binary image, the "pixel block corresponding to the pixel coordinate is labeled as a hole in the labeled image according to the pixel coordinate of the real hole in the updated foreground region", that is, the "pixel coordinate of the real hole in the updated binary image" is labeled as a hole in the labeled image.
In one possible real-time manner, when the foreground region is the foreground region in the binary image, after the pixel block corresponding to the pixel coordinate is marked as a hole in the marked image, whether the pixel block marked as the hole in the marked image really belongs to the hole part can be further determined by a manual checking manner.
In this embodiment, an annotation instruction input by the user may be received. Wherein, the labeling instruction comprises: and the object corresponding to the target pixel block in the pixel blocks marked as the holes in the marked image. The target object refers to a pixel block marked as a hole in the marked image, but in reality, the pixel block corresponds to a non-hole. For example, the object corresponding to the target pixel block is a road, a house, or the like.
In this scenario, in this embodiment, the noted image and the pixel blocks noted as holes in the noted image are displayed, and in this embodiment, the user may select the target pixel block by clicking or other selecting means, so as to trigger sending of the noted instruction to the image noted device. The labeling instruction comprises the following steps: and an object corresponding to the target pixel block. Correspondingly, the image labeling device receives a labeling instruction input by a user, and can label the target pixel block as an object corresponding to the target pixel block according to the labeling instruction in the labeled image, so that the labeling accuracy of the pixel block can be improved.
The image labeling method provided in the embodiment comprises the following steps: generating a foreground region corresponding to the target object according to the pixel coordinates of the target object in the marked image; generating a hypothesis hole in the foreground region, and updating the foreground region according to a judging result of whether the hypothesis hole is a real hole, wherein the updated foreground region comprises the real hole; and marking the pixel blocks corresponding to the pixel coordinates as holes in the marked image according to the pixel coordinates of the real holes in the updated foreground region. The image labeling method provided by the embodiment can determine the true cavity in the foreground area labeled as the target object, so that the cavity in the target object can be labeled again, and the accuracy of image labeling is improved.
On the basis of the above embodiment, the image labeling method provided by the present application is further described below with reference to fig. 5 by taking a foreground region as an example in a binary image, and fig. 5 is a second schematic flowchart of the image labeling method provided by the present application. As shown in fig. 5, the method provided in this embodiment may include:
s501, generating a binary image according to pixel coordinates of a target object in the marked image, wherein a foreground area of the binary image is the target object.
In this embodiment, the hypothesis hole may be generated multiple times, and the foreground area of the binary image is updated according to the determination result of whether the hypothesis hole generated each time is a real hole, and the updated binary image is obtained after multiple updating processes. This process is described in detail below in connection with S502-S506. It should be understood that the purpose of generating the hypothesis hole multiple times in this embodiment is that the pixel block of the foreground region may be completely traversed for the purpose of generating the hypothesis hole, so as to accurately obtain the real hole in the foreground region.
S502, in the binary image P corresponding to the iteration period i i The method comprises the steps of randomly generating a hypothesis hole with a preset probability, wherein i is an integer greater than or equal to 1.
The binary image generated by the marked image is taken as a binary image P corresponding to an iteration period 1 1 . It will be appreciated that in the binary image P corresponding to iteration cycle 1 1 The judgment result of whether the assumed cavity is a real cavity is that the binary image P corresponding to the iteration period 1 1 The binary image obtained after updating the foreground region is a binary image P corresponding to the iteration period 2 2 。
In this embodiment, the binary image P corresponding to the iteration period i can be obtained with a preset probability i Is used to randomly generate hypothesis holes. i is an integer greater than or equal to 1. Wherein, in each iteration period i, the corresponding binary image P i The method for randomly generating the hypothesis holes in the foreground region of (a) may refer to the description of S102 in the above embodiment, and will not be described herein.
S503, according to the pixel coordinates of the assumed holes, determining P in the marked image according to the corresponding relation between the distribution of the assumed holes and the distribution of the real holes i Whether the hypothetical hole on the foreground region of (a) is a true hole.
In this embodiment, a hypothetical hole is obtained in the foreground image of the binary image, and the pixel coordinates of the hypothetical hole can be determined correspondingly. Since the binary image has the same size as the annotated image, the position of the hypothetical hole in the annotated image (i.e. the pixel coordinates) can be determined from the pixel coordinates of the hypothetical hole in the annotated image. According to the corresponding relation between the distribution of the assumed holes and the distribution of the real holes in the marked image, namely according to the CRF algorithm (the CRF algorithm is used for representing the corresponding relation between the distribution of the assumed holes and the distribution of the real holes), whether the assumed holes in the marked image are the real holes or not can be determined, and corresponding, P can be determined i Whether the hypothetical hole on the foreground region of (a) is a true hole.
S504, according to P i For P, the judgment result of whether the assumed cavity on the foreground region is a true cavity i Marking, corroding and expanding in sequence to obtain an updated binary image P corresponding to the iteration period i i ”。
In the present embodiment, P is acquired i After judging whether the assumed cavity on the foreground region is a true cavity, P can be determined as follows i And (5) processing. Wherein the treatment is to sequentially treat P i Marking, corrosion and expansion treatment are sequentially carried out, and then an updated binary image P corresponding to the iteration period i can be obtained i ”。
Alternatively, in this embodiment, P may be i The pixel block corresponding to the assumed hole which is the real hole on the foreground area is marked as the background area, and P is marked i The pixel blocks corresponding to the assumed holes which are not real holes on the foreground region are marked as the foreground region, and a marked binary image corresponding to the iteration period i is obtained, namely the real holes in the binary image are re-marked.
Further, in this embodiment, the erosion process is performed on the labeled binary image corresponding to the iteration cycle i, so as to obtain an erosion processed binary image P corresponding to the iteration cycle i i '. Wherein the etching treatment is performed to reduce the prospectThe area of the region is actually aimed at connecting adjacent tiny holes into large holes, so as to avoid that after the next iteration period further generates a hypothetical hole, the tiny hole is re-marked as a foreground region, and therefore idle work is performed. In this embodiment, the adjacent micro holes are connected to form a large hole, so that the real hole marked in the iteration period i has higher stability.
Further, in this embodiment, the binary image P after the etching process corresponding to the iteration period i may be further i ' expansion treatment to obtain P i ". Wherein, since the etching treatment is performed before, the corresponding marked contour area of the foreground region is reduced, and the expansion treatment is performed to restore the corresponding marked contour area of the foreground region to the point before the etching treatment is not performed. It should be understood that the expansion factor and corrosion factor in this step are the same in this embodiment.
S505, judging whether i is equal to a preset value, if so, executing E, otherwise, executing P i "as the binary image corresponding to the next iteration cycle, and return to execute A.
In this embodiment, a preset value of the iteration period may be preset, so that the randomly generated hypothesis hole may be expected to traverse the pixel block corresponding to the foreground region. If i is determined to be the preset value, determining that iteration is ended, and updating the binary image P corresponding to the iteration period i i "as an updated binary image. If it is determined that i is not the preset value, it is determined that iteration is further required to be continued, i is added with 1, and P is added i And 'as a binary image corresponding to the next iteration period', returning to execute S502 until i is a preset value.
S506, the updated binary image P corresponding to the iteration period i i "as an updated binary image.
In this embodiment, when the iteration is ended, that is, when the iteration period i is a preset value, the updated binary image P corresponding to the iteration period i is obtained i "as an updated binary image.
S507, obtaining pixel coordinates of a real hole in the updated binary image by comparing a foreground region corresponding to the first iteration period with a foreground region corresponding to the last iteration period, wherein the pixel coordinates of the real hole are as follows: in the first iteration cycle, the pixel coordinates corresponding to the pixel blocks belonging to the foreground region in the last iteration cycle.
At the end of the iteration, the foreground region corresponding to the first iteration period and the foreground region corresponding to the last iteration period are compared, so that the pixel coordinates corresponding to the pixel blocks belonging to the foreground region in the first iteration period and the background region in the last iteration period can be determined. The pixel coordinates are the pixel coordinates of the real holes in the updated binary image. Therefore, in this embodiment, the pixel coordinates of the real hole in the updated binary image can be obtained by comparing the foreground region corresponding to the first iteration period with the foreground region corresponding to the last iteration period.
Illustratively, the foreground region corresponding to the first iteration period does not include a real hole, that is, the pixel blocks in the foreground region all belong to the target object. In the last iteration period, the pixel block which belongs to the foreground region in the original mode is marked as a real hole, and the pixel block A which is marked as the background region is the real hole, and the corresponding pixel coordinate of the pixel block A is the pixel coordinate of the real hole.
Alternatively, in this embodiment, the pixel coordinates corresponding to the pixel blocks belonging to the background area in the first iteration period and the foreground area in the last iteration period may be determined by comparing the background area corresponding to the first iteration period with the background area corresponding to the last iteration period. The pixel block corresponding to the pixel coordinates is: the labeled image is originally labeled as a background region, but should be labeled as a block of pixels of the target object.
Correspondingly, in this embodiment, the pixel coordinate may be obtained, and the pixel block corresponding to the pixel coordinate is marked as the target object in the marked image.
In this embodiment, the assumed holes may be generated multiple times in the foreground area of the binary image, and the foreground area of the binary image is updated according to the determination result of whether the assumed holes generated each time are real holes, so as to obtain the updated binary image after multiple times of processing. The purpose of generating the hypothesis hole for multiple times is that the generated hypothesis hole can completely traverse the pixel blocks of the foreground region, so that the true hole in the foreground region can be accurately acquired. Further, after determining the real hole of the foreground region in each iteration period, the foreground region can be subjected to corrosion and expansion processing, so that the adjacent tiny holes are connected into large holes, and the situation that the tiny holes are re-marked as the foreground region after further generating the hypothesized holes in the next iteration period is avoided, so that the real holes marked in the iteration period have higher stability, and the accuracy of hole marking is improved.
Fig. 6 is a schematic structural diagram of an image labeling device provided by the application. The image marking device can be electronic equipment such as a server or a terminal (such as a smart phone, a tablet computer, a computer and the like). As shown in fig. 6, the image labeling apparatus 600 includes: a first processing module 601, a second processing module 602 and a transceiver module 603.
The first processing module 601 is configured to generate a foreground area corresponding to the target object according to the pixel coordinates of the target object in the labeled image, generate a hypothetical hole in the foreground area, and update the foreground area according to a determination result of whether the hypothetical hole is a real hole, where the updated foreground area includes the real hole.
The second processing module 602 is configured to label, in the labeled image, a pixel block corresponding to the pixel coordinate as a hole according to the pixel coordinate of the real hole in the updated foreground area.
Optionally, the foreground region is a foreground region in a binary image.
Correspondingly, the first processing module 601 is specifically configured to generate a hypothetical hole in a foreground area in the binary image, and update the binary image according to a determination result of whether the hypothetical hole is a real hole, where the foreground area in the updated binary image is an updated foreground area.
And a second processing module 602, configured to label, in the labeled image, the pixel block corresponding to the pixel coordinate as a hole according to the pixel coordinate of the real hole in the updated binary image.
Optionally, the first processing module 601 is specifically configured to:
A. binary image P corresponding to iteration period i i Randomly generating a hypothesis cavity according to a preset probability, wherein i is an integer greater than or equal to 1;
B. judging P according to the corresponding relation between the distribution of the assumed holes and the distribution of the real holes in the marked image according to the pixel coordinates of the assumed holes i Whether the assumed hole on the foreground region is a real hole;
C. according to P i For P, the judgment result of whether the assumed cavity on the foreground region is a true cavity i Marking, corroding and expanding in sequence to obtain an updated binary image P corresponding to the iteration period i i ”；
D. Judging whether i is equal to a preset value, if so, executing E, and if not, executing P i "as the binary image corresponding to the next iteration period, and return to execute A;
E. the updated binary image P corresponding to the iteration period i i "as an updated binary image.
Optionally, the first processing module 601 is specifically configured to: will P i The pixel block corresponding to the assumed hole which is the real hole on the foreground area is marked as the background area, and P is marked i The pixel blocks corresponding to the assumed holes which are not real holes on the foreground region are marked as the foreground region, and a marked binary image corresponding to the iteration period i is obtained; carrying out corrosion treatment on the marked binary image corresponding to the iteration period i to obtain a corrosion treated binary image P corresponding to the iteration period i i 'A'; the corresponding corrosion-treated binary image P of iteration period i i ' expansion treatment to obtain P i ”。
Optionally, the first processing module 601 is further configured to obtain pixel coordinates of a real hole in the updated binary image by comparing a foreground area corresponding to the first iteration period with a foreground area corresponding to the last iteration period, where the pixel coordinates of the real hole are: in the first iteration cycle, the pixel coordinates corresponding to the pixel blocks belonging to the foreground region in the last iteration cycle.
Optionally, the image labeling device further includes: a transceiver module 603.
The transceiver module 603 is configured to receive a first instruction input by a user before the foreground area generates the hypothesis hole, where the first instruction is used to instruct expansion processing on the foreground area of the binary image, and the first instruction is input when a labeling contour area of the target object in the labeled image is smaller than a first preset contour area; and/or the number of the groups of groups,
The second instruction is used for indicating to perform corrosion treatment on the foreground area of the binary image, and the second instruction is as follows: and inputting when the labeling contour area of the target object in the labeled image is larger than a second preset contour area, wherein the second preset contour area is larger than the first preset contour area.
Optionally, the transceiver module 603 is further configured to receive a labeling instruction input by a user, where the labeling instruction includes: and the object corresponding to the target pixel block in the pixel blocks marked as the holes in the marked image.
Correspondingly, the second processing module 602 is further configured to label the target pixel block as an object corresponding to the target pixel block according to the labeling instruction.
The principle and technical effects of the image labeling device provided in this embodiment are similar to those of the image labeling method, and are not described herein.
Fig. 7 is a schematic structural diagram of an electronic device provided by the present application. As shown in fig. 7, the electronic device 700 includes: a memory 701 and at least one processor 702.
Memory 701 for storing program instructions.
The processor 702 is configured to implement the image labeling method in this embodiment when the program instructions are executed, and the specific implementation principle can be seen from the above embodiment, which is not described herein again.
The electronic device 700 may also include and input/output interface 703.
The input/output interface 703 may include a separate output interface and an input interface, or may be an integrated interface that integrates input and output. The output interface is used for outputting data, and the input interface is used for acquiring the input data.
The present application also provides a readable storage medium having stored therein execution instructions which, when executed by at least one processor of an electronic device, when executed by the processor, implement the image labeling method in the above embodiment.
The present application also provides a program product comprising execution instructions stored in a readable storage medium. The at least one processor of the electronic device may read the execution instructions from the readable storage medium, and execution of the execution instructions by the at least one processor causes the electronic device to implement the image annotation methods provided by the various embodiments described above.
In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in hardware plus software functional units.
The integrated units implemented in the form of software functional units described above may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium, and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (english: processor) to perform some of the steps of the methods according to the embodiments of the application. And the aforementioned storage medium includes: u disk, mobile hard disk, read-Only Memory (ROM), random access Memory (Random Access Memory, RAM), magnetic disk or optical disk, etc.
In the above embodiment of the electronic device, it should be understood that the processor may be a central processing unit (english: central Processing Unit, abbreviated as CPU), or may be other general purpose processors, digital signal processors (english: digital Signal Processor, abbreviated as DSP), application specific integrated circuits (english: application Specific Integrated Circuit, abbreviated as ASIC), or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the present application may be embodied directly in a hardware processor for execution, or in a combination of hardware and software modules in a processor for execution.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the application.
Claims (10)
1. An image labeling method, comprising:
generating a foreground region corresponding to a target object in the marked image according to the pixel coordinates of the target object;
generating a hypothesis hole in the foreground region, updating the foreground region according to a judging result of whether the hypothesis hole is a real hole, wherein the updated foreground region comprises the real hole;
and marking the pixel blocks corresponding to the pixel coordinates as holes in the marked image according to the pixel coordinates of the real holes in the updated foreground region.
2. The method according to claim 1, wherein the foreground region is a foreground region in a binary image corresponding to the noted image, the generating a hypothetical hole in the foreground region, and updating the foreground region according to a determination result of whether the hypothetical hole is a real hole, includes:
generating a hypothesis hole in a foreground region in the binary image, and updating the binary image according to a judging result of whether the hypothesis hole is a real hole, wherein the foreground region in the updated binary image is the updated foreground region;
and marking the pixel block corresponding to the pixel coordinates as a hole in the marked image according to the pixel coordinates of the real hole in the updated foreground region, wherein the marking comprises the following steps:
And marking the pixel block corresponding to the pixel coordinates as a hole in the marked image according to the pixel coordinates of the real hole in the updated binary image.
3. The method according to claim 2, wherein generating a hypothetical hole in the foreground region of the binary image and updating the binary image according to a determination result of whether the hypothetical hole is a true hole comprises:
A. binary image P corresponding to iteration period i i Randomly generating a hypothesis cavity according to a preset probability, wherein i is an integer greater than or equal to 1;
B. judging the P in the marked image according to the pixel coordinates of the assumed holes and the corresponding relation between the distribution of the assumed holes and the distribution of the real holes i Whether the assumed hole on the foreground region is a real hole;
C. according to the P i Judging whether the assumed cavity on the foreground region is a true cavity or not according to the judgment result of the true cavity, and judging the P i Marking, corroding and expanding in sequence to obtain an updated binary image P corresponding to the iteration period i i ”；
D. Judging whether the i is equal to a preset value, if so, executing E, and if not, executing the P i "as the binary image corresponding to the next iteration period, and executing the A in a return manner;
E. The updated binary image P corresponding to the iteration period i is processed i "as the updated binary image.
4. A method according to claim 3, wherein said P is defined by i Judging whether the assumed cavity on the foreground region is a true cavity or not according to the judgment result of the true cavity, and judging the P i Marking, corroding and expanding in sequence to obtain an updated binary image P corresponding to the iteration period i i ", comprising:
the P is set i The pixel block corresponding to the assumed hole which is the real hole on the foreground area is marked as the background area, and the P is marked i The pixel blocks corresponding to the assumed holes which are not real holes on the foreground region are marked as the foreground region, and a marked binary image corresponding to the iteration period i is obtained;
performing corrosion treatment on the marked binary image corresponding to the iteration period i to obtain a corrosion treated binary image P corresponding to the iteration period i i '；
A binary image P corresponding to the iteration period i after corrosion treatment i ' performing expansion treatment to obtain the P i ”。
5. The method according to claim 4, wherein before labeling the pixel block corresponding to the pixel coordinates as the hole in the labeled image according to the pixel coordinates of the real hole in the updated binary image, the method further comprises:
Obtaining pixel coordinates of a real hole in the updated binary image by comparing a foreground area corresponding to a first iteration period with a foreground area corresponding to a last iteration period, wherein the pixel coordinates of the real hole are as follows: in the first iteration period, the pixel coordinates corresponding to the pixel blocks belonging to the foreground region in the last iteration period.
6. The method of claim 2, wherein before generating the hypothesis hole in the foreground region in the binary image, further comprises:
receiving a first instruction input by a user, wherein the first instruction is used for indicating to expand a foreground region of the binary image, and is input when the labeling contour area of a target object in the labeled image is smaller than a first preset contour area; and/or the number of the groups of groups,
and receiving a second instruction input by the user, wherein the second instruction is used for indicating the foreground region of the binary image to be corroded, and the second instruction is input when the marked contour area of the target object in the marked image is larger than a second preset contour area, and the second preset contour area is larger than the first preset contour area.
7. The method according to claim 2, wherein after labeling the pixel block corresponding to the pixel coordinates as a hole in the labeled image according to the pixel coordinates of the real hole in the updated binary image, the method further comprises:
receiving an annotation instruction input by a user, wherein the annotation instruction comprises the following steps: an object corresponding to a target pixel block in the pixel blocks marked as holes in the marked image;
and marking the target pixel block as an object corresponding to the target pixel block according to the marking instruction.
8. An image marking apparatus, comprising:
the first processing module is used for generating a foreground area corresponding to the target object according to the pixel coordinates of the target object in the marked image; generating a hypothesis hole in the foreground region, updating the foreground region according to the judgment result of whether the hypothesis hole is a real hole, and processing the updated foreground region, wherein the updated foreground region comprises the real hole;
and the second processing module is used for marking the pixel block corresponding to the pixel coordinates as a hole in the marked image according to the pixel coordinates of the real hole in the updated foreground region.
9. An electronic device, comprising: at least one processor and memory;
the memory stores computer-executable instructions;
the at least one processor executing computer-executable instructions stored in the memory to cause the electronic device to perform the method of any one of claims 1-7.
10. A computer readable storage medium having stored thereon computer executable instructions which, when executed by a processor, implement the method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910814719.1A CN112529830B (en) | 2019-08-30 | 2019-08-30 | Image labeling method, device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910814719.1A CN112529830B (en) | 2019-08-30 | 2019-08-30 | Image labeling method, device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112529830A CN112529830A (en) | 2021-03-19 |
| CN112529830B true CN112529830B (en) | 2023-11-14 |
Family
ID=74974098
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910814719.1A Active CN112529830B (en) | 2019-08-30 | 2019-08-30 | Image labeling method, device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112529830B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1104916A1 (en) * | 1999-12-04 | 2001-06-06 | Luratech Gesellschaft für Luft-und Raumfahrt-Technologie & Multimedia mbH | Method for compressing color and/or grey-level scanned documents |
| CN107909138A (en) * | 2017-11-14 | 2018-04-13 | 江苏大学 | A kind of class rounded grain thing method of counting based on Android platform |
| CN108961246A (en) * | 2018-07-10 | 2018-12-07 | 吉林大学 | A kind of scanning electron microscope image hole recognition methods based on artificial intelligence |
| CN109308456A (en) * | 2018-08-31 | 2019-02-05 | 北京字节跳动网络技术有限公司 | The information of target object determines method, apparatus, equipment and storage medium |
| CN109583444A (en) * | 2018-11-22 | 2019-04-05 | 博志生物科技有限公司 | Hole region localization method, device and computer readable storage medium |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20130073459A (en) * | 2011-12-23 | 2013-07-03 | 삼성전자주식회사 | Method and apparatus for generating multi-view |
| US10032092B2 (en) * | 2016-02-02 | 2018-07-24 | Adobe Systems Incorporated | Training data to increase pixel labeling accuracy |
| EP3396370B1 (en) * | 2017-04-28 | 2021-11-17 | Fujitsu Limited | Detecting portions of interest in images |
-
2019
- 2019-08-30 CN CN201910814719.1A patent/CN112529830B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1104916A1 (en) * | 1999-12-04 | 2001-06-06 | Luratech Gesellschaft für Luft-und Raumfahrt-Technologie & Multimedia mbH | Method for compressing color and/or grey-level scanned documents |
| CN107909138A (en) * | 2017-11-14 | 2018-04-13 | 江苏大学 | A kind of class rounded grain thing method of counting based on Android platform |
| CN108961246A (en) * | 2018-07-10 | 2018-12-07 | 吉林大学 | A kind of scanning electron microscope image hole recognition methods based on artificial intelligence |
| CN109308456A (en) * | 2018-08-31 | 2019-02-05 | 北京字节跳动网络技术有限公司 | The information of target object determines method, apparatus, equipment and storage medium |
| CN109583444A (en) * | 2018-11-22 | 2019-04-05 | 博志生物科技有限公司 | Hole region localization method, device and computer readable storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 新闻视频静态图形标识分割;王建;贺翼虎;周源华;;上海交通大学学报(第05期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112529830A (en) | 2021-03-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112580623B (en) | Image generation method, model training method, related device and electronic equipment | |
| CN109840477B (en) | Method and device for recognizing shielded face based on feature transformation | |
| CN114186632B (en) | Method, device, equipment and storage medium for training key point detection model | |
| CN113705461B (en) | Face definition detection method, device, equipment and storage medium | |
| CN110008997B (en) | Image texture similarity recognition method, device and computer readable storage medium | |
| CN112651953B (en) | Picture similarity calculation method and device, computer equipment and storage medium | |
| CN112419132B (en) | Video watermark detection method, device, electronic equipment and storage medium | |
| CN115620321B (en) | Table identification method and device, electronic equipment and storage medium | |
| CN115100659B (en) | Text recognition method, device, electronic equipment and storage medium | |
| CN114581646A (en) | Text recognition method and device, electronic equipment and storage medium | |
| CN114494058B (en) | Image processing method, device, electronic equipment and storage medium | |
| CN115861255A (en) | Model training method, device, equipment, medium and product for image processing | |
| CN115101069A (en) | Voice control method, device, equipment, storage medium and program product | |
| CN113379592B (en) | Processing method and device for sensitive area in picture and electronic equipment | |
| CN112529830B (en) | Image labeling method, device, electronic equipment and storage medium | |
| CN116468985B (en) | Model training method, quality detection device, electronic equipment and medium | |
| CN117216591A (en) | Training method and device for three-dimensional model matching and multi-modal feature mapping model | |
| CN110414845B (en) | Risk assessment method and device for target transaction | |
| CN115439850B (en) | Method, device, equipment and storage medium for identifying image-text characters based on examination sheets | |
| CN115018734B (en) | Video restoration method and training method and device of video restoration model | |
| CN113128696A (en) | Distributed machine learning communication optimization method and device, server and terminal equipment | |
| CN116166583A (en) | Data precision conversion method and device, DMA controller and medium | |
| CN114463361A (en) | Network model training method, device, equipment, medium and program product | |
| CN114359645B (en) | Image expansion method, device, equipment and storage medium based on characteristic area | |
| CN113688645B (en) | Identification method, system and equipment |
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 |