CN114913321A - Object attention mining method and system based on local-to-global knowledge migration - Google Patents
Object attention mining method and system based on local-to-global knowledge migration Download PDFInfo
- Publication number
- CN114913321A CN114913321A CN202210324744.3A CN202210324744A CN114913321A CN 114913321 A CN114913321 A CN 114913321A CN 202210324744 A CN202210324744 A CN 202210324744A CN 114913321 A CN114913321 A CN 114913321A
- Authority
- CN
- China
- Prior art keywords
- image
- attention
- local
- global
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the technical field of image processing, and provides an object attention mining method and system based on local to global knowledge migration. By generating the attention image under multiple visual angles, non-salient regions in more images can be found, by designing a knowledge migration loss function, the attention information under the local visual angle can be effectively migrated to a global model, and the generated object attention region can play a role in various weak supervision semantic segmentation tasks.
Description
Technical Field
The invention belongs to the technical field of image processing, and particularly relates to an object attention mining method and system based on local to global knowledge migration.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
In recent years, the deep learning algorithm brings rapid development to the semantic segmentation task. However, training a deep neural network for semantic segmentation requires a large number of pixel-level labels, which consumes a lot of manpower and resources. Therefore, in order to reduce the dependency on pixel-level labels, there have been many approaches in recent years that attempt to study weakly supervised semantic segmentation based on image-level labels. If the stages are used for distinguishing, the weak supervision semantics can be divided into a single stage and a double stage. In single-stage weak supervised semantic segmentation, an end-to-end segmented network is generally trained by directly using image-level labels as supervision. Early approaches viewed this as a multi-instance learning problem. Later, George pandandreou proposed an expectation maximization approach, with intermediate prediction results to supervise semantic segmentation networks. Bingfeng Zhang et al in "relevance multiple mat: An end-to-end well hyper detailed segmentation approach" uses image classification branches for generating attention images and constructing pseudo segmentation labels, thereby supervising another parallel segmentation branch. The method of Single-stage segmentation from image labels proposes to generate pseudo labels by using the appearance prior characteristics of images during training. In addition, Jianjun Chen et al, in the "End-to-End boundary extension for well-featured segmentation" method, construct an End-to-End frame with the encoder-decoder network for finding the image boundaries.
Although the single-stage weak supervised semantic segmentation method has achieved a certain success, there is still a gap from the performance point of view from the double-stage weak supervised semantic segmentation method. The two-stage weak supervision semantic segmentation method utilizes an attention image to generate segmentation pseudo labels, and then utilizes the pseudo labels to train a segmentation network. Thus, the core of this approach is to generate high quality attention images. Yunchao Wei et al propose in the "Object region mining with adaptive operation: A simple classification to segmentation approach" method, based on a strategy of anti-erasure, to drive a classification network to discover a new Object region by iteratively erasing the excavated Object region. Qibin Hou et al propose a Self-erasing strategy to prevent the attention area from spreading into the background in the "Self-erasing network for integral object attack" approach. After that, Alexander Kolesnikov et al proposed the idea of Seed diffusion in "Seed, expanded and constraint: Three principles for well-visualized image segmentation". Later, Peng-Tao Jiang et al proposed a strategy for on-line accumulation of attention images at different training phases in "Integral object mining visual on-line assessment". Yu-Ting Chang et al explores subcategory information in "week-super detailed segmentation sub-category expansion" to emphasize non-salient object regions.
In addition, there are efforts to refine the attention image with accurate boundaries to obtain a complete object region. Jiwwoon Ahn et al, in "Learning pixel-level detailed with image-level super vision for image super vision segmentation" propagates attention-seeking to strongly responding areas on an image by Learning inter-pixel similarities. Seungho Lee et al in "Railroad is not a train: Saliency as pseudo-pixel persistence for the road persistence segmentation" use the generated Saliency image as a region supervision motivates the model to generate a high quality attention image.
The above methods all complete the task of refining the attention image under the global view angle, and have the following problems: under a single global view angle, the classification network cannot capture more object detail information, only can dig less non-significant object areas, and the generated target object areas are not obvious or cannot accurately extract the target object areas.
Disclosure of Invention
In order to solve at least one technical problem existing in the background art, the invention provides an object attention mining method and system based on local-to-global knowledge migration, which are provided with a local model and a global model, wherein the local model is used for extracting an attention area in a local image block, and the global model extracts a more complete object attention image by learning detailed knowledge from the local model; .
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides an object attention mining method based on local-to-global knowledge migration, which comprises the following steps:
acquiring a global image to be excavated;
obtaining a target object region according to the global image to be excavated and the object attention excavation model;
the construction process of the object attention mining model comprises the following steps: on the basis of a local model and a global model, randomly dividing a global image to be excavated into a plurality of local image blocks, and obtaining an attention image under multiple viewing angles according to the plurality of local image blocks and the trained local model; and obtaining a complete attention image based on the global image to be mined, the attention image under the multi-view and the trained global model.
A second aspect of the present invention provides an object attention mining system based on local-to-global knowledge migration, comprising:
an image acquisition module configured to: acquiring a global image to be excavated;
a target object region acquisition module configured to: obtaining a target object region according to the global image to be excavated and the object attention excavation model;
the construction process of the object attention mining model comprises the following steps: constructing a local model and a global model, randomly dividing a global image to be excavated into a plurality of local image blocks, and obtaining an attention image under multiple visual angles according to the plurality of local image blocks and the trained local model; and obtaining a complete attention image based on the global image to be mined, the attention image under the multi-view and the trained global model.
A third aspect of the invention provides a computer-readable storage medium.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method for object attention mining based on local-to-global knowledge migration as described above.
A fourth aspect of the invention provides a computer apparatus.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor when executing the program implementing the steps in the local-to-global knowledge migration based object attention mining method as described above.
Compared with the prior art, the invention has the beneficial effects that:
the invention sets two models of local and global, wherein the local model is used to extract the attention area in the local image block, and the global model extracts a more complete object attention image by learning detailed knowledge from the local model.
According to the method, the non-significant region in more images can be found by generating the attention image under multiple visual angles, the attention information under the local visual angle can be effectively transferred to the global model by designing a knowledge transfer loss function, and the generated object attention region can play a role in various weak supervision semantic segmentation tasks.
Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
FIG. 1 is a schematic diagram of an object attention mining method based on local-to-global knowledge migration;
FIG. 2 is a schematic diagram of an object attention mining method based on local-to-global knowledge migration and migration loss functions;
3(a) -3 (d) are comparative graphs of attention generated by different methods under the first image;
4(a) -4 (d) are comparative graphs of attention generated by different methods under a second image;
FIGS. 5(a) -5 (d) are comparative graphs of attention generated by different methods under a third image;
6(a) -6 (g) are weak supervised semantic segmentation effect graphs of different methods under a fourth image;
7(a) -7 (g) are weak supervised semantic segmentation effect graphs of different methods under a fifth image;
8(a) -8 (g) are weak supervised semantic segmentation effect graphs of different methods under a sixth image;
fig. 9(a) -9 (g) are weak supervised semantic segmentation effect graphs of different methods under a seventh image.
Detailed Description
The invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Example one
As shown in fig. 1, the present embodiment provides an object attention mining method based on local-to-global knowledge migration, including the following steps:
s101, acquiring a global image to be mined;
s102, obtaining a target object region according to the global image to be excavated and the object attention excavation model;
the construction process of the object attention mining model comprises the following steps: based on a local model and a global model, randomly dividing a global image to be excavated into a plurality of local image blocks, and obtaining an attention image under multiple visual angles according to the plurality of local image blocks and the trained local model; and obtaining a complete attention image based on the global image to be mined, the attention image under the multi-view and the trained global model.
The technical scheme has the advantages that the local model is used for extracting the attention area in the local image block, and the global model extracts a complete object attention image by learning detailed knowledge from the local model; according to the method, the global image is divided into the attention images under multiple visual angles, so that the method can be used for exploring non-significant areas in more images.
As one or more embodiments, attention migration and shape migration loss functions are introduced into the global model training process.
The process of training the global model by adopting the attention migration and shape migration loss functions comprises the following steps:
by optimizing the L2 loss function of the corresponding region between the local attention image and the global attention image activated by the Softmax function, the object detail information learned under the local view angle can be migrated to the global network, and the process is attention migration. On the basis, the invention also uses the generated saliency image and the binarized local attention image to carry out element-by-element multiplication, and then calculates the loss function between the pixel value and the global attention. This process, shape migration, aims to migrate the exact shape information contained in the saliency image onto the global network together.
As one or more embodiments, the invention designs a knowledge migration loss function, which ensures that the attention information under the local view angle can be effectively migrated to the global model.
The attention migration loss function is:
wherein N is the number of local views, A i Represents the ith local attention image, G i And showing the intercepted part of the global attention image after the Softmax is activated on the partial area corresponding to the ith local view angle.
The shape migration loss function is:
wherein N is the number of local views, A i Representing the ith local attention image, B i Representing A after binarization i ，S i Representing a portion of the saliency image taken at the i-th local perspective corresponding region, G i And showing the intercepted part of the global attention image after the Softmax is activated on the ith local view angle corresponding area.
The method has the advantages that in the training process, the local model takes a plurality of local image blocks as input, the global model takes the whole image as input, and the attention image which is generated by the local model and contains rich detail information is used as supervision to train the global model by introducing two loss functions of attention migration and shape migration; the loss function of the supervised global model can migrate the shape information of the attention image and the saliency image extracted by the local model from the plurality of local images into the global network simultaneously.
As shown in fig. 1, when training the local model and the global model, the local network takes image blocks of different randomly cut regions in the image as input, and the global network takes the whole image as input.
After forward calculation, the local network outputs the attention images of the local image blocks, and the local attention images are used for training the global network to learn detailed knowledge in each region, so that the knowledge of the local network is migrated to the global network, and the global network can extract a more complete attention image.
As one or more embodiments, as shown in fig. 2, when the global model is trained, it is determined whether a target object region is a saliency image, if so, the attention image under multiple views is binarized and then multiplied by the saliency image element by element, and the multiplied image result is used as shape and detail information to supervise the global model, otherwise, the attention image under multiple views is directly used for supervision training.
The technical scheme aims to transfer the shape information in the saliency image to a global network, and in addition, because all pictures do not have saliency objects, when a saliency region does not exist, the local attention image is directly used for supervision, and a recognition result can be obtained more accurately according to the existence of the saliency objects.
As shown in fig. 3(a) -3 (d), 4(a) -4 (d), and 5(a) -5 (d), which are comparison diagrams of attention images generated by different methods under three images, a category activation image (Local) generated by a Local network, a category activation image (L2G) adopting a Local-to-global model, and a category activation image (L2G w/shape) adopting a Local-to-global and shape migration are compared, and it can be seen from the comparison diagrams that a target object region obtained by the shape migration category activation image (L2G w/shape) method adopted in the present invention can be extracted more finely and completely.
As shown in fig. 6(a) -6 (g), 7(a) -7 (g), 8(a) -8 (g) and 9(a) -9 (g), the weakly supervised semantic segmentation effect maps of the four different methods under different images are represented by segmentation labels (CAM) generated by a general network, segmentation labels (Local) generated by a Local network, segmentation labels (L2G) generated by a Local-to-global framework, segmentation labels (Local + Shape) generated by a Local network using a saliency image for refinement, segmentation labels (L2G + Shape) generated by a Local-to-global framework using Shape migration and real segmentation labels (GT) manually labeled.
Wherein the average cross-over ratio over the extended validation set of VOCs 2012 and the authentic tags is: general networks (CAM) 47.5%, Local networks (Local) 50.0%, Local to global skeleton (L2G) 54.9%, segmentation tags for Local networks using saliency image refinement (L2G + Shape) 69.9%, segmentation tags for Local to global skeleton generated using Shape migration (L2G + Shape) 72.1%.
By comparing the segmentation effects, the method provided by the invention extracts the learning detail knowledge to extract a relatively complete object attention image without introducing other targets, so that the accuracy is highest.
Example two
The present embodiment provides an object attention mining system based on local-to-global knowledge migration, including:
an image acquisition module configured to: acquiring a global image to be excavated;
a target object region acquisition module configured to: obtaining a target object region according to the global image to be excavated and the object attention excavation model;
the construction process of the object attention mining model comprises the following steps: constructing a local model and a global model, randomly dividing a global image to be excavated into a plurality of local image blocks, and obtaining an attention image under multiple visual angles according to the plurality of local image blocks and the trained local model; and obtaining a complete attention image based on the global image to be mined, the attention image under the multi-view and the trained global model.
Attention migration and shape migration loss functions are introduced in the global model training process.
When the global model is trained, whether a target object area is a saliency image is judged, if yes, the attention image under multiple visual angles is subjected to binarization processing and then is subjected to element-by-element multiplication with the saliency image, and an image result obtained by multiplication is used as shape and detail information to supervise the global model, otherwise, the attention image under multiple visual angles is directly adopted to supervise and train.
EXAMPLE III
The present embodiments provide a computer-readable storage medium having stored thereon a computer program which, when being executed by a processor, carries out the steps of the method for attention mining of objects based on local-to-global knowledge migration as described above.
Example four
The present embodiment provides a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps in the local-to-global knowledge migration based object attention mining method as described above when executing the program.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above may be implemented by a computer program, which may be stored in a computer readable storage medium and executed by a computer to implement the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. The method for mining the attention of the object based on the local-to-global knowledge migration is characterized by comprising the following steps of:
acquiring a global image to be excavated;
obtaining a target object region according to the global image to be excavated and the object attention excavation model;
the construction process of the object attention mining model comprises the following steps: on the basis of a local model and a global model, randomly dividing a global image to be excavated into a plurality of local image blocks, and obtaining an attention image under multiple viewing angles according to the plurality of local image blocks and the trained local model; and obtaining a complete attention image based on the global image to be mined, the attention image under the multi-view and the trained global model.
2. The method for object attention mining based on local-to-global knowledge migration according to claim 1, wherein attention migration and shape migration loss functions are introduced in the global model training process.
3. The method for object attention mining based on local-to-global knowledge migration according to claim 2, wherein the process of training the global model using the attention migration loss function comprises:
and migrating the learned object detail information under the local view onto the global network by optimizing the loss function of the corresponding region between the local attention image and the global attention image activated by the Softmax function.
4. The method for object attention mining based on local-to-global knowledge migration according to claim 2, wherein the process of training the global model using the shape migration loss function comprises: and judging whether the target object area is a saliency image, if so, performing binarization processing on the attention image under multiple viewing angles, performing element-by-element multiplication on the processed saliency image and the saliency image, and supervising the global model by taking the image result obtained by the multiplication as shape and detail information, otherwise, directly performing supervision training by adopting the attention image under the multiple viewing angles.
5. The method of object attention mining based on local-to-global knowledge migration according to claim 2, wherein the attention migration function is:
wherein N is the number of local views, A i Representing the ith local attention image, G i The part of the global attention image after the Softmax is activated, which is intercepted on the partial area corresponding to the ith local visual angle, is represented;
the shape migration loss function is:
wherein N is the number of local views, A i Representing the ith local attention image, B i Representing A after binarization i ，S i Represent a significant image inPart of the ith local view angle corresponding to the region, G i And showing the intercepted part of the global attention image after the Softmax is activated on the ith local view angle corresponding area.
6. An object attention mining system based on local-to-global knowledge migration, comprising:
an image acquisition module configured to: acquiring a global image to be excavated;
a target object region acquisition module configured to: obtaining a target object region according to the global image to be excavated and the object attention excavation model;
the construction process of the object attention mining model comprises the following steps: on the basis of a local model and a global model, randomly dividing a global image to be excavated into a plurality of local image blocks, and obtaining an attention image under multiple viewing angles according to the plurality of local image blocks and the trained local model; and obtaining a complete attention image based on the global image to be mined, the attention image under the multi-view angle and the trained global model.
7. The local-to-global knowledge migration based object attention mining system of claim 6, wherein attention migration and shape migration loss functions are introduced into the global model training process.
8. The local-to-global knowledge migration based object attention mining system of claim 6, wherein the process of training the global model using the shape migration loss function comprises: and judging whether the target object area is a saliency image, if so, performing binarization processing on the attention image under multiple viewing angles, performing element-by-element multiplication on the processed saliency image and the saliency image, and supervising the global model by taking the image result obtained by the multiplication as shape and detail information, otherwise, directly performing supervision training by adopting the attention image under the multiple viewing angles.
9. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method for attention mining of objects based on local-to-global knowledge migration according to any one of claims 1 to 5.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements the steps in the method for object attention mining based on local-to-global knowledge migration of any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210324744.3A CN114913321A (en) | 2022-03-30 | 2022-03-30 | Object attention mining method and system based on local-to-global knowledge migration |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210324744.3A CN114913321A (en) | 2022-03-30 | 2022-03-30 | Object attention mining method and system based on local-to-global knowledge migration |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114913321A true CN114913321A (en) | 2022-08-16 |
Family
ID=82763765
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210324744.3A Pending CN114913321A (en) | 2022-03-30 | 2022-03-30 | Object attention mining method and system based on local-to-global knowledge migration |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114913321A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115277264A (en) * | 2022-09-28 | 2022-11-01 | 季华实验室 | Subtitle generating method based on federal learning, electronic equipment and storage medium |
| CN115393598A (en) * | 2022-10-31 | 2022-11-25 | 南京理工大学 | Weakly supervised semantic segmentation method based on non-salient region object mining |
-
2022
- 2022-03-30 CN CN202210324744.3A patent/CN114913321A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115277264A (en) * | 2022-09-28 | 2022-11-01 | 季华实验室 | Subtitle generating method based on federal learning, electronic equipment and storage medium |
| CN115393598A (en) * | 2022-10-31 | 2022-11-25 | 南京理工大学 | Weakly supervised semantic segmentation method based on non-salient region object mining |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10229346B1 (en) | Learning method, learning device for detecting object using edge image and testing method, testing device using the same | |
| CN111986099B (en) | Tillage monitoring method and system based on convolutional neural network with residual error correction fused | |
| Qiao et al. | LGPMA: complicated table structure recognition with local and global pyramid mask alignment | |
| CN102592268B (en) | Method for segmenting foreground image | |
| CN106570453B (en) | Method, device and system for pedestrian detection | |
| CN106951830B (en) | Image scene multi-object marking method based on prior condition constraint | |
| CN114913321A (en) | Object attention mining method and system based on local-to-global knowledge migration | |
| US20140270489A1 (en) | Learned mid-level representation for contour and object detection | |
| Montoya-Zegarra et al. | Semantic segmentation of aerial images in urban areas with class-specific higher-order cliques | |
| CN112418216B (en) | Text detection method in complex natural scene image | |
| CN102831239B (en) | A kind of method and system building image data base | |
| Bowen et al. | Algorithmic identification of looted archaeological sites from space | |
| CN114418021B (en) | Model optimization method, device and computer program product | |
| CN101908214A (en) | Moving object detection method with background reconstruction based on neighborhood correlation | |
| Hernández et al. | Shape ultimate attribute opening | |
| CN115909079A (en) | Crack detection method combining depth feature and self-attention model and related equipment | |
| CN112801109A (en) | Remote sensing image segmentation method and system based on multi-scale feature fusion | |
| Abujayyab et al. | Integrating object-based and pixel-based segmentation for building footprint extraction from satellite images | |
| CN114512052B (en) | Method and device for generating diverging and converging intersections by fusing remote sensing images and track data | |
| Luo et al. | AD-RoadNet: an auxiliary-decoding road extraction network improving connectivity while preserving multiscale road details | |
| KR102026280B1 (en) | Method and system for scene text detection using deep learning | |
| Callier et al. | Automatic road area extraction from printed maps based on linear feature detection | |
| Ai et al. | Geometry preserving active polygon-incorporated sign detection algorithm | |
| CN113609980A (en) | Lane line sensing method and device for automatic driving vehicle | |
| CN112614134A (en) | Image segmentation method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |