CN114821681A - Fingerprint augmentation method - Google Patents
Fingerprint augmentation method Download PDFInfo
- Publication number
- CN114821681A CN114821681A CN202210736029.0A CN202210736029A CN114821681A CN 114821681 A CN114821681 A CN 114821681A CN 202210736029 A CN202210736029 A CN 202210736029A CN 114821681 A CN114821681 A CN 114821681A
- Authority
- CN
- China
- Prior art keywords
- fingerprint image
- fingerprint
- loss
- normal
- backward
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/12—Fingerprints or palmprints
- G06V40/1347—Preprocessing; Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
Abstract
The invention discloses a fingerprint augmentation method, which comprises the following steps: collecting a normal fingerprint image and a variable fingerprint image of the same fingerprint to construct a training data set; training a cyclic countermeasure network according to a training data set to ensure that the generated fingerprint image and the input fingerprint image have the same identity information; and inputting the fingerprint image to be changed into the trained loop countermeasure network to generate an expected change fingerprint image. Training a circulating countermeasure network through the actually acquired normal fingerprint image and the actually acquired change fingerprint image, so that the circulating countermeasure network can map the normal fingerprint image into the change fingerprint image; and it is ensured by the cyclic loss that the changed fingerprint image has the same identity information as the normal fingerprint image.
Description
Technical Field
The invention relates to the technical field of fingerprint identification, in particular to a fingerprint augmentation method.
Background
Fingerprinting algorithms currently require more effort in addressing fingerprint variations. With the wide application of machine learning algorithms, the role of fingerprint data is more and more obvious. However, since the fingerprint data can carry personal identity information and is often used for authentication, the fingerprint data is not easy to collect as sensitive information. It is necessary to transform the fingerprint image, which currently carries identity information, as necessary to simulate the many changes that may occur. These variations include, but are not limited to: physiological changes such as sloughing, deepening or transition of folds, sudden trauma leading to scars, etc. These newly generated fingerprint images need to approximate as realistically as possible the changes that actually occur (given by the sample fingerprint pair) and ensure that the identity information of the fingerprint images is not lost, which is not addressed by currently known techniques.
Disclosure of Invention
Therefore, the technical problem to be solved by the present invention is to overcome the defect that the physiological change of the fingerprint cannot be simulated in the prior art, so as to provide a fingerprint augmentation method, which can simulate the physiological change of the fingerprint and ensure that the identity information of the fingerprint is not lost, and particularly to a fingerprint augmentation method.
The invention provides a fingerprint augmentation method, which comprises the following steps:
s1: collecting a normal fingerprint image and a variable fingerprint image of the same fingerprint to construct a training data set;
s2: training a cyclic countermeasure network according to a training data set to ensure that the generated fingerprint image and the input fingerprint image have the same identity information;
s3: and inputting the fingerprint image to be changed into the trained loop countermeasure network to generate an expected change fingerprint image.
Preferably, in S1, constructing the training data set includes: collecting fingerprint images of the same fingerprint in a normal state, and recording the fingerprint images as a first normal fingerprint image; collecting fingerprint images under the same fingerprint physiological change state, and recording the fingerprint images as first change fingerprint images; forming a training fingerprint image pair by using a first normal fingerprint image and a first variable fingerprint image of the same fingerprint; and constructing a training data set according to the training fingerprint image pairs corresponding to the plurality of fingerprints.
Preferably, the loop countermeasure network includes a generator and an arbiter; the generator comprises a first generator and a second generator, wherein the first generator is used for transforming the input normal fingerprint image, and the second generator is used for transforming the input variable fingerprint image; the discriminator comprises a first discriminator and a second discriminator, and the first discriminator and the second discriminator are used for obtaining the discrimination loss.
Preferably, in S2, the process of training the loop countermeasure network includes a forward loop process and a backward loop process, and the forward loop process and the backward loop process jointly obtain the loop loss and the discrimination loss; the cyclic loss comprises forward cyclic loss and backward cyclic loss, and the discrimination loss comprises first discrimination loss and second discrimination loss; obtaining forward circulation loss and first discrimination loss through a forward circulation process; obtaining backward circulation loss and second judgment loss through a backward circulation process; through the cyclic loss and the discrimination loss, the generated change fingerprint image is an expected change fingerprint image, and the expected change fingerprint image retains the same identity information as the fingerprint image to be changed.
Preferably, the forward loop process includes: inputting the first normal fingerprint image into a first generator to obtain a forward variation fingerprint image, and inputting the forward variation fingerprint image into a second generator to obtain a forward normal fingerprint image; calculating forward circulation loss according to the first normal fingerprint image and the forward normal fingerprint image; and inputting the first normal fingerprint image and the forward normal fingerprint image into a first discriminator to obtain a first discrimination loss.
Preferably, the specific content of the backward circulation process is as follows: inputting the first change fingerprint image into a second generator to obtain a backward normal fingerprint image, and inputting the backward normal fingerprint image into the first generator to obtain a backward change fingerprint image; calculating backward cyclic loss according to the first variation fingerprint image and the backward variation fingerprint image; and inputting the first change fingerprint image and the backward change fingerprint image into a second discriminator to obtain a second discrimination loss.
Preferably, the cyclic loss includes a perceptual loss and a pixel loss;
the process of obtaining the perception loss is as follows: inputting the fingerprint image and the transformed fingerprint image into a fingerprint feature extraction network to obtain corresponding features of the fingerprint image and corresponding features of the transformed fingerprint image, and taking the Euclidean distance between the two corresponding features as a sensing loss; the fingerprint feature extraction network is a pre-trained convolutional neural network;
the process of obtaining the pixel loss is as follows: and taking the Euclidean distance between the fingerprint image and the transformed fingerprint image as pixel loss.
Preferably, the discrimination process of the discriminator is as follows: inputting the fingerprint image and the transformed fingerprint image into a discriminator, and judging the fingerprint image as true and the transformed fingerprint image as fake when the discriminator judges that the fingerprint image is normal; when the discriminator judges the error, the mean square error loss function is adopted as the discrimination loss.
Preferably, when the fingerprint image is the first normal fingerprint image, the transformed fingerprint image is a forward normal fingerprint image; when the fingerprint image is a first change fingerprint image, the transformed fingerprint image is a backward change fingerprint image.
Preferably, the physiological change state comprises moulting, wrinkling, or scarring.
The technical scheme of the invention has the following advantages: training a circulating countermeasure network through the actually acquired normal fingerprint image and the actually acquired change fingerprint image, so that the circulating countermeasure network can map the normal fingerprint image into the change fingerprint image; and ensure that the changed fingerprint image has the same identity information as the normal fingerprint image.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flow chart of a fingerprint augmentation method in accordance with an embodiment of the present invention;
FIG. 2 is a schematic diagram of a forward loop process of a fingerprint augmentation method according to an embodiment of the present invention;
FIG. 3 is a diagram illustrating a backward loop process of a fingerprint augmentation method according to an embodiment of the present invention.
Detailed Description
The technical solutions of the present invention will be described clearly and completely with reference to the accompanying drawings, and it should be understood that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In addition, the technical features involved in the different embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
As shown in fig. 1, the present embodiment provides a fingerprint augmentation method, including:
s1: collecting a normal fingerprint image and a variable fingerprint image of the same fingerprint to construct a training data set;
specifically, fingerprint images of the same fingerprint in a normal state are collected and recorded as a first normal fingerprint image; collecting fingerprint images under the same fingerprint physiological change state, and recording the fingerprint images as first change fingerprint images; a training fingerprint image pair is formed by a first normal fingerprint image and a first variant fingerprint image of the same fingerprint, for example: (fingerprint image, fingerprint image at exuviation), (fingerprint image, wrinkle fingerprint image), (fingerprint image, scar fingerprint image), (wrinkle fingerprint image, fingerprint image after wiping skin care product); constructing a training data set according to training fingerprint image pairs corresponding to the plurality of fingerprints; the physiological change state includes molting, wrinkling (deepening or shallowing), scar, etc.
S2: training a cyclic countermeasure network according to a training data set to ensure that the generated fingerprint image and the input fingerprint image have the same identity information;
specifically, a cyclic countermeasure network is trained according to a normal fingerprint image and a changed fingerprint image in a training data set, so that the cyclic countermeasure network learns a mapping relation, and the mapping relation is used for mapping the normal fingerprint image into the changed fingerprint image; through the cyclic loss and the discrimination loss obtained in the training process, the change fingerprint image is an expected change fingerprint image, and the identity information of the change fingerprint image is kept the same as that of a normal fingerprint image;
a generator and a discriminator are arranged in the loop countermeasure network, the generator comprises a first generator and a second generator, the first generator is used for transforming the input normal fingerprint image, and the second generator is used for transforming the input change fingerprint image; the discriminator comprises a first discriminator and a second discriminator, and the first discriminator and the second discriminator are used for obtaining discrimination loss.
The process of training the cyclic countermeasure network comprises a forward cycle process and a backward cycle process, and the forward cycle process and the backward cycle process jointly obtain cycle loss and discrimination loss; the cyclic loss comprises forward cyclic loss and backward cyclic loss, and the discrimination loss comprises first discrimination loss and second discrimination loss; obtaining forward circulation loss and first discrimination loss through a forward circulation process; obtaining backward circulation loss and second judgment loss through a backward circulation process; through the cyclic loss and the discrimination loss, the generated change fingerprint image is an expected change fingerprint image, and the expected change fingerprint image retains the same identity information as the fingerprint image to be changed.
As shown in fig. 2, the specific content of the forward loop process is: inputting a first normal fingerprint image into a first generator to obtain a forward variation fingerprint image, and inputting the forward variation fingerprint image into a second generator to obtain a forward normal fingerprint image; calculating forward circulation loss according to the first normal fingerprint image and the forward normal fingerprint image; and inputting the first normal fingerprint image and the forward normal fingerprint image into a first discriminator to obtain a first discrimination loss.
As shown in fig. 3, the specific content of the backward loop process is: inputting the first change fingerprint image into a second generator to obtain a backward normal fingerprint image, and inputting the backward normal fingerprint image into the first generator to obtain a backward change fingerprint image; calculating backward cyclic loss according to the first variation fingerprint image and the backward variation fingerprint image; and inputting the first change fingerprint image and the backward change fingerprint image into a second discriminator to obtain a second discrimination loss.
And in the whole training process, the whole loop countermeasure network is continuously optimized through the loop loss and the discriminant loss. Wherein the cyclic losses include perceptual losses and pixel losses.
Taking the forward loop process as an example:
the process of obtaining the perception loss is as follows: inputting the first normal fingerprint image and the forward normal fingerprint image into a fingerprint feature extraction network to obtain corresponding features of the first normal fingerprint image and corresponding features of the forward normal fingerprint image, wherein the fingerprint feature extraction network is a pre-trained convolutional neural network, and a MobileFaceNet network is adopted in the embodiment; since the convolutional neural network can extract higher-level information (such as identity class) of the fingerprint image, the euclidean distance between two corresponding features is used as the perceptual loss in this embodiment.
The process of obtaining the pixel loss is as follows: in this embodiment, the euclidean distance between the first normal fingerprint image and the forward normal fingerprint image is calculated as the pixel loss. The pixels lose more information at lower levels of the image of interest (pixel values of the image).
The discrimination process of the discriminator is as follows: inputting the first normal fingerprint image and the forward normal fingerprint image into a discriminator, wherein when the discriminator judges that the fingerprint image is normal, the collected first normal fingerprint image is judged to be true, and the forward normal fingerprint image is judged to be fake; when the discriminator judges the error, the present embodiment adopts the mean square error loss function number as the discrimination loss, and the discrimination loss is not 0, and the optimization of the discriminator is known in the process of gradient back propagation through the mean square error loss function.
Specifically, the forward loop process is taken as an example: when the first normal fingerprint image and the forward normal fingerprint image are input into the first discriminator, two corresponding vectors are obtained, and the discriminator considers that in an ideal state, after the first normal fingerprint image is input into the first discriminator, a vector which is all 1, namely (1, 1, 1, 1, …, 1, 1, 1) is obtained, and after the forward normal fingerprint image is input into the first discriminator, a vector which is all 0 is obtained. However, in the training process, if the vector obtained after the first normal fingerprint image is input into the first discriminator is not a vector with all 1, then a Mean Squared Error (MSE) is performed on the vector with all 1; similarly, the vector obtained after the forward normal fingerprint is input into the first discriminator is not a vector with all 0, and the mean square error is carried out on the vector with all 0; the sum of the two mean square errors is used as the first discriminant loss.
When the loop countermeasure network is optimized through the loop loss and the discriminant loss, the discriminator hopes to accurately judge the authenticity of the input fingerprint image, and the generator hopes to generate a good fingerprint image so as to 'cheat' the discriminator, which is a maximum and minimum game, and the optimized target Nash equilibrium is reached when the training is finished, so that the first generator can estimate the distribution of the inverted normal fingerprint.
S3: and inputting the fingerprint image to be changed into the trained loop countermeasure network to generate an expected change fingerprint image.
The cyclic countermeasure network examines two cyclic errors in a normal fingerprint image domain and a changed fingerprint image domain to ensure that the changed fingerprint image and the normal fingerprint image have the same identity information; meanwhile, the loop countermeasure network can ensure that the change fingerprint image obtained by mapping is the expected change fingerprint image through the first discriminator and the second discriminator.
The fingerprint augmentation method provided by the embodiment has the following beneficial effects: training a circulating countermeasure network through the actually acquired normal fingerprint image and the actually acquired change fingerprint image, so that the circulating countermeasure network can map the normal fingerprint image into the change fingerprint image; and it is ensured by the cyclic loss that the changed fingerprint image has the same identity information as the normal fingerprint image.
It should be understood that the above examples are only for clarity of illustration and are not intended to limit the embodiments. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. And obvious variations or modifications therefrom are within the scope of the invention.
Claims (10)
1. A fingerprint augmentation method, comprising:
s1: collecting a normal fingerprint image and a variable fingerprint image of the same fingerprint to construct a training data set;
s2: training a cyclic countermeasure network according to a training data set to ensure that the generated fingerprint image and the input fingerprint image have the same identity information;
s3: and inputting the fingerprint image to be changed into the trained loop countermeasure network to generate an expected change fingerprint image.
2. The fingerprint augmentation method of claim 1, wherein in S1, constructing the training data set comprises: collecting fingerprint images of the same fingerprint in a normal state, and recording the fingerprint images as a first normal fingerprint image; collecting fingerprint images under the same fingerprint physiological change state, and recording the fingerprint images as first change fingerprint images; forming a training fingerprint image pair by using a first normal fingerprint image and a first variable fingerprint image of the same fingerprint; and constructing a training data set according to the training fingerprint image pairs corresponding to the plurality of fingerprints.
3. The fingerprint augmentation method of claim 2, wherein the loop countermeasure network comprises a generator and an arbiter; the generator comprises a first generator and a second generator, wherein the first generator is used for transforming the input normal fingerprint image, and the second generator is used for transforming the input variable fingerprint image; the discriminator comprises a first discriminator and a second discriminator, and the first discriminator and the second discriminator are used for obtaining the discrimination loss.
4. The fingerprint augmentation method of claim 3, wherein in the step S2, the process of training the loop countermeasure network includes a forward loop process and a backward loop process, and the forward loop process and the backward loop process jointly obtain the loop loss and the discrimination loss; the cyclic loss comprises a forward cyclic loss and a backward cyclic loss, and the discrimination loss comprises a first discrimination loss and a second discrimination loss; obtaining forward circulation loss and first discrimination loss through a forward circulation process; obtaining backward circulation loss and second judgment loss through a backward circulation process; through the cyclic loss and the discrimination loss, the generated change fingerprint image is an expected change fingerprint image, and the expected change fingerprint image retains the same identity information as the fingerprint image to be changed.
5. The fingerprint augmentation method of claim 4, wherein the forward loop process comprises: inputting the first normal fingerprint image into a first generator to obtain a forward variation fingerprint image, and inputting the forward variation fingerprint image into a second generator to obtain a forward normal fingerprint image; calculating forward circulation loss according to the first normal fingerprint image and the forward normal fingerprint image; and inputting the first normal fingerprint image and the forward normal fingerprint image into a first discriminator to obtain a first discrimination loss.
6. The fingerprint augmentation method of claim 5, wherein the backward loop process comprises: inputting the first change fingerprint image into a second generator to obtain a backward normal fingerprint image, and inputting the backward normal fingerprint image into the first generator to obtain a backward change fingerprint image; calculating backward cyclic loss according to the first variation fingerprint image and the backward variation fingerprint image; and inputting the first change fingerprint image and the backward change fingerprint image into a second discriminator to obtain a second discrimination loss.
7. The fingerprint augmentation method of claim 6, wherein the cyclic losses include perceptual losses and pixel losses;
the process of obtaining the perception loss is as follows: inputting the fingerprint image and the transformed fingerprint image into a fingerprint feature extraction network to obtain corresponding features of the fingerprint image and corresponding features of the transformed fingerprint image, and taking the Euclidean distance between the two corresponding features as a sensing loss; the fingerprint feature extraction network is a pre-trained convolutional neural network;
the process of obtaining the pixel loss is as follows: and taking the Euclidean distance between the fingerprint image and the transformed fingerprint image as pixel loss.
8. The fingerprint augmentation method of claim 7, wherein the discriminator comprises: inputting the fingerprint image and the transformed fingerprint image into a discriminator, and judging the fingerprint image as true and the transformed fingerprint image as fake when the discriminator judges that the fingerprint image is normal; when the discriminator judges the error, the mean square error loss function is adopted as the discrimination loss.
9. The fingerprint augmentation method of claim 8, wherein when the fingerprint image is the first normal fingerprint image, the transformed fingerprint image is a forward normal fingerprint image; when the fingerprint image is a first change fingerprint image, the transformed fingerprint image is a backward change fingerprint image.
10. The method of claim 2, wherein the physiological change comprises moulting, wrinkling, or bruising.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210736029.0A CN114821681A (en) | 2022-06-27 | 2022-06-27 | Fingerprint augmentation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210736029.0A CN114821681A (en) | 2022-06-27 | 2022-06-27 | Fingerprint augmentation method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114821681A true CN114821681A (en) | 2022-07-29 |
Family
ID=82522856
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210736029.0A Pending CN114821681A (en) | 2022-06-27 | 2022-06-27 | Fingerprint augmentation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114821681A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2622032A (en) * | 2022-08-31 | 2024-03-06 | Bluestar Software Ltd | Automated footwear identification system and method |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108470187A (en) * | 2018-02-26 | 2018-08-31 | 华南理工大学 | A kind of class imbalance question classification method based on expansion training dataset |
| CN109978792A (en) * | 2019-03-28 | 2019-07-05 | 厦门美图之家科技有限公司 | A method of generating image enhancement model |
| CN110147830A (en) * | 2019-05-07 | 2019-08-20 | 东软集团股份有限公司 | Training image data generates method, image data classification method and the device of network |
| CN110415309A (en) * | 2019-06-26 | 2019-11-05 | 公安部第三研究所 | The method automatically generated based on generation confrontation network implementations fingerprint picture |
| CN112270648A (en) * | 2020-09-24 | 2021-01-26 | 清华大学 | Unsupervised image transformation method and unsupervised image transformation device based on loop countermeasure network |
| WO2021116036A1 (en) * | 2019-12-12 | 2021-06-17 | Koninklijke Philips N.V. | Data augmentation for a machine learning method |
| CN114140662A (en) * | 2021-11-29 | 2022-03-04 | 广州中科智巡科技有限公司 | Insulator lightning stroke image sample amplification method based on cyclic generation countermeasure network |
| CN114463789A (en) * | 2022-01-19 | 2022-05-10 | 北京至简墨奇科技有限公司 | Non-contact fingerprint image enhancement method, apparatus, storage medium and program product |
| CN114565946A (en) * | 2022-01-25 | 2022-05-31 | 西安电子科技大学 | Fingerprint living body detection method based on self-attention mechanism lightweight network |
| US20220180518A1 (en) * | 2019-03-08 | 2022-06-09 | University Of Southern California | Improved histopathology classification through machine self-learning of "tissue fingerprints" |
-
2022
- 2022-06-27 CN CN202210736029.0A patent/CN114821681A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108470187A (en) * | 2018-02-26 | 2018-08-31 | 华南理工大学 | A kind of class imbalance question classification method based on expansion training dataset |
| US20220180518A1 (en) * | 2019-03-08 | 2022-06-09 | University Of Southern California | Improved histopathology classification through machine self-learning of "tissue fingerprints" |
| CN109978792A (en) * | 2019-03-28 | 2019-07-05 | 厦门美图之家科技有限公司 | A method of generating image enhancement model |
| CN110147830A (en) * | 2019-05-07 | 2019-08-20 | 东软集团股份有限公司 | Training image data generates method, image data classification method and the device of network |
| CN110415309A (en) * | 2019-06-26 | 2019-11-05 | 公安部第三研究所 | The method automatically generated based on generation confrontation network implementations fingerprint picture |
| WO2021116036A1 (en) * | 2019-12-12 | 2021-06-17 | Koninklijke Philips N.V. | Data augmentation for a machine learning method |
| CN112270648A (en) * | 2020-09-24 | 2021-01-26 | 清华大学 | Unsupervised image transformation method and unsupervised image transformation device based on loop countermeasure network |
| CN114140662A (en) * | 2021-11-29 | 2022-03-04 | 广州中科智巡科技有限公司 | Insulator lightning stroke image sample amplification method based on cyclic generation countermeasure network |
| CN114463789A (en) * | 2022-01-19 | 2022-05-10 | 北京至简墨奇科技有限公司 | Non-contact fingerprint image enhancement method, apparatus, storage medium and program product |
| CN114565946A (en) * | 2022-01-25 | 2022-05-31 | 西安电子科技大学 | Fingerprint living body detection method based on self-attention mechanism lightweight network |
Non-Patent Citations (3)
| Title |
|---|
| AI TAKAHASHI 等: "Domain Transformation of Fingerprint Images Using CycleGAN", 《2019 IEEE 8TH GLOBAL CONFERENCE ON CONSUMER ELECTRONICS (GCCE)》 * |
| SHERVIN MINAEE 等: "Finger-GAN: Generating Realistic Fingerprint Images Using Connectivity Imposed GAN", 《ARXIV》 * |
| 耿鑫: "基于生成对抗网络的去雾算法研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2622032A (en) * | 2022-08-31 | 2024-03-06 | Bluestar Software Ltd | Automated footwear identification system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liao et al. | Deep facial spatiotemporal network for engagement prediction in online learning | |
| Liu et al. | Contactless palmprint identification using deeply learned residual features | |
| CN106778468B (en) | 3D face identification method and equipment | |
| Jan et al. | Accurate facial parts localization and deep learning for 3D facial expression recognition | |
| CN106446754A (en) | Image identification method, metric learning method, image source identification method and devices | |
| CN109657595A (en) | Based on the key feature Region Matching face identification method for stacking hourglass network | |
| CN106778474A (en) | 3D human body recognition methods and equipment | |
| Emeršič et al. | The unconstrained ear recognition challenge 2019 | |
| CN106778489A (en) | The method for building up and equipment of face 3D characteristic identity information banks | |
| CN110119695A (en) | A kind of iris activity test method based on Fusion Features and machine learning | |
| CN112132099A (en) | Identity recognition method, palm print key point detection model training method and device | |
| CN111507184B (en) | Human body posture detection method based on parallel cavity convolution and body structure constraint | |
| CN106529377A (en) | Age estimating method, age estimating device and age estimating system based on image | |
| CN113570497A (en) | Image processing method, image processing device, computer equipment and storage medium | |
| Lakshmi et al. | Off-line signature verification using Neural Networks | |
| CN114821681A (en) | Fingerprint augmentation method | |
| CN106778491B (en) | The acquisition methods and equipment of face 3D characteristic information | |
| Daas et al. | A review of finger vein biometrics authentication system | |
| CN112016592B (en) | Domain adaptive semantic segmentation method and device based on cross domain category perception | |
| CN108304800A (en) | A kind of method of Face datection and face alignment | |
| CN113095156A (en) | Double-current network signature identification method and device based on inverse gray scale mode | |
| Liu | RETRACTED ARTICLE: Video Face Detection Based on Deep Learning | |
| CN109886212A (en) | From the method and apparatus of rolling fingerprint synthesis fingerprint on site | |
| CN114360058A (en) | Cross-visual angle gait recognition method based on walking visual angle prediction | |
| Khadilkar et al. | Face identification based on discrete wavelet transform and neural networks |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220729 |
|
| RJ01 | Rejection of invention patent application after publication |