EP4165605A1 - Procédé de génération d'images d'une caméra intérieure de véhicule - Google Patents
Procédé de génération d'images d'une caméra intérieure de véhiculeInfo
- Publication number
- EP4165605A1 EP4165605A1 EP21730593.7A EP21730593A EP4165605A1 EP 4165605 A1 EP4165605 A1 EP 4165605A1 EP 21730593 A EP21730593 A EP 21730593A EP 4165605 A1 EP4165605 A1 EP 4165605A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- individual
- image
- vehicle
- images
- 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
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000009826 distribution Methods 0.000 claims abstract description 14
- 238000010200 validation analysis Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 206010041349 Somnolence Diseases 0.000 description 1
- 210000005069 ears Anatomy 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/10—Geometric effects
- G06T15/20—Perspective computation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T13/00—Animation
- G06T13/20—3D [Three Dimensional] animation
- G06T13/40—3D [Three Dimensional] animation of characters, e.g. humans, animals or virtual beings
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/50—Lighting effects
- G06T15/506—Illumination models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- 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
- G06V10/776—Validation; Performance evaluation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/59—Context or environment of the image inside of a vehicle, e.g. relating to seat occupancy, driver state or inner lighting conditions
Definitions
- the present application relates to a method for generating synthetic images, simulating images of individuals acquired by an interior vehicle camera.
- the application finds an advantageous application in the learning or validation of algorithms for processing images acquired by a vehicle interior camera.
- Machine learning algorithms require large amounts of training data to be reliable. This is particularly the case for algorithms based on the processing of images acquired by an interior vehicle camera, and oriented towards a driver or a passenger of the vehicle, and which can have various applications such as detection of characteristic points of the face. , detection of drowsiness or distraction, etc.
- training data that is to say images acquired by an interior vehicle camera, which covers a sufficiently large variety of possible acquisition conditions, because this implies making drive different models of vehicles, under different conditions, with different passengers or drivers, etc.
- the training database should also include extreme cases, whatever the parameter considered (user position, brightness, level of blur, etc.), that can be difficult to obtain.
- the constitution of a training database can be extremely time-consuming because each image acquired by a vehicle interior camera must then be manually annotated to identify characteristic points of the image. By considering a manual annotation time of the order of 2 to 4 minutes per image, we can easily measure the cost and the time required. annotation of an image database that may include tens or even hundreds of thousands of images.
- the aim of the invention is to improve the situation.
- an aim of the invention is to allow the constitution of a database of images acquired by a vehicle interior camera which is simpler and less expensive.
- Another object of the invention is to obtain an image database also including extreme cases for the variable parameters.
- the invention provides a method for generating synthetic images, each image simulating an image of an individual acquired by an interior vehicle camera, the method being implemented by a computer and comprising:
- each model comprising a three-dimensional representation of an individual's head
- each configuration corresponding to a combination of values or states taken by each parameter, so that the set of configurations is representative of the probability distribution of each parameter
- each image corresponds to a configuration generated for a variable parameter, and where each image further includes the three-dimensional positions of a set of characteristic points of the individual's head, and
- the method further comprises the generation of several vehicle models, and the generation of the images is implemented for each individual model in each vehicle model.
- the method comprises generating, for each individual model, a set of images comprising an image for each of the configurations generated for each of the variable parameters received.
- the method further comprises receiving a number N of configurations per individual model, and generating, for each individual model, a set of configurations comprising a total of N configurations for all the variable parameters.
- the method comprises receiving at least one variable parameter relating to the camera from the group consisting of:
- the method comprises receiving at least one variable parameter relating to a position of the individual relative to a seat of the vehicle and / or at least one variable parameter relating to the environment of the vehicle. model of the individual, including:
- the method further comprises:
- the subject of the invention is also a method for learning or validating an algorithm based on processing of an image acquired by an interior vehicle camera, comprising:
- Another subject of the invention is the use of a database of images generated by the implementation of the method according to the above description for the learning or validation of an algorithm based on a processing. of an image acquired by a vehicle interior camera.
- the invention also relates to a computer program product, comprising code instructions for implementing the methods according to the above description, when executed by a computer.
- the invention finally relates to a device for generating synthetic images simulating images of individuals acquired by an interior vehicle camera, comprising a computer and a memory, in which the computer is configured for the implementation. methods according to the above description.
- the proposed invention makes it possible to artificially generate an image database where each image simulates an image taken by an interior vehicle camera.
- the images of the image database represent several models of individuals, in variable contexts obtained by varying different parameters according to distribution probabilities. This allows the image database to be representative of all the configurations and their probabilities of occurrence.
- the images generated already include, thanks to the models of individuals which are three-dimensional models of faces, three-dimensional positions of characteristic points of the head of the individual represented. It is therefore not necessary to carry out a manual annotation of each image, which represents a significant saving of time.
- FIG. 1 schematically represents an example of implementation of a method for generating synthetic images.
- FIG. 2 represents an example of a synthetic image generated by the implementation of the method.
- a computer 1 which may for example be a processor, a microprocessor, a controller, a microcontroller, etc.
- This method makes it possible to quickly generate a large quantity of synthetic images, where each synthetic image represents the head of an individual in a vehicle, and simulates an image which would have been taken of the individual by an interior camera. vehicle.
- the images are generated by varying numerous parameters, in order to enrich the image database thus obtained.
- the method comprises a generation 100 of a plurality of models of individuals, also called avatars.
- Each individual model includes a three-dimensional representation of an individual's head. More specifically, each individual model can include a three-dimensional point cloud, corresponding to the contours of the head and face, and a texture applied to the point cloud, simulating the appearance of the individual's head, that is to say an image the points of which are associated with a color or a level of intensity allowing, once applied to the contours of the head, to give the head a human appearance.
- At least ten, and preferably several dozen different individual models are generated.
- the models of individuals are generated by varying a set of parameters comprising all or part of the following parameters:
- the method also includes generating 110 multiple vehicle models, each vehicle model including a three-dimensional representation of the interior of a vehicle cabin. This makes it possible to generate images of the models of individuals in the various vehicle models.
- the method then includes receiving 200 of a set of variable parameters and a probability distribution associated with each parameter.
- the parameters can be selected by a user. It is understood by "variable parameter" that each parameter can take a plurality of values or states.
- a variable parameter can correspond to a continuous or discrete quantity.
- the parameters received relate to at least one of the environment of the individual model, the pose of the individual, that is to say the orientation of the head of the individual, or the position of the individual relative to his environment, that is to say to the vehicle in which he is located, and more particularly the position of the head of the individual relative to the seat of the vehicle in which it is located.
- the parameters relating to the environment of the individual model may include parameters relating to the interior camera of the vehicle, including in particular:
- the level of blur which may include a level of blur associated with each shot of the camera (foreground where the individual is located and background),
- the parameters relating to the environment of the individual model can also include the environment outside the vehicle, which is visible through the windows of the vehicle which can appear on an image acquired by an interior camera, depending on its position and orientation .
- the parameters relating to the environment of the model of the individual may also include the light intensity and / or the direction of the light illuminating the scene acquired by the interior camera. These parameters are notably variable depending on the time of day or night, and the meteorological conditions considered.
- the probability distribution associated with each parameter can also be selected or configured by the user.
- the probability distributions can for example be Gaussian or uniform as a function of the parameters considered.
- the method can also comprise the reception 210 of a movement or of a series of movements that each model of individual must achieve, which can be indicated by the user.
- This step can also be implemented during the step 100 for generating the avatar model, if it is the avatar generation algorithm which has this functionality.
- the movement or sequence of movements to be performed for an individual can be defined by an initial position (for example orientation of the head of the individual according to three angles), a final position, a speed of movement between the initial position and the final position.
- Step 210 may also include receiving the number of times the movement has been performed in a video.
- the method then comprises the generation 300 of a set of configurations, each configuration corresponding to a combination of values or states taken by each of the variable parameters received.
- a first configuration can include:
- a second configuration can include:
- the configurations are generated by varying each parameter according to its associated probability distribution, so that the set of configurations is representative of the probability distribution of each parameter.
- the set of configurations obtained can for example take the form of a configuration file, where each configuration is defined by the value or the state taken by each parameter.
- the method comprises the reception 220 of a number N of configurations to be generated, and the generation of the configurations therefore comprises the generation of a total of N configurations, for all of the variable parameters received.
- the number N can be defined by the user as a function of the quantity of images which he seeks to produce at the end of the method, which also depends on the number of models of individuals and of the number of models of vehicle.
- the method then comprises the generation 400, for each individual model, of a set of images simulating images of the individual model acquired by a camera inside the vehicle, where each image corresponds to one of the generated configurations.
- the method comprises generating, for each individual model, and for each vehicle model, a set of images comprising an image for each of the generated configurations. A total number of images is therefore obtained equal to the number N of configurations, multiplied by the number of models of individuals, or multiplied by the number of models of individuals and the number of models of vehicle.
- each individual model comprises a set of three-dimensional points corresponding to the characteristic points of the head and of the face, the positions of these points are known, which allows each image generated to also include the positions said characteristic points of the individual's head on the image.
- An example of an image where the positions of characteristic points of the individual's face are highlighted is shown in figure 2. In this way, there is no need to manually annotate the images for later use. .
- the method can comprise, for at least one configuration, or for several, or even all of the configurations, the generation 410 of a video representing the series of movements performed by the individual model, and the generation of all the images that compose it.
- the method comprises recording 500, in memory 2, of all of the images (and the positions of the associated characteristic points) and of the videos, so as to form a rich database comprising several models of 'individuals, in several vehicle models, and with very varied parameters, including in particular extreme values of parameters.
- this database can then be used directly for the learning, or the validation, of an algorithm based on a processing of an image acquired by an interior vehicle camera, in particular an algorithm of machine learning.
- the user can determine the variable parameters useful to provide for the implementation of the method.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer Graphics (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Geometry (AREA)
- Image Processing (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
FR2006233A FR3111460B1 (fr) | 2020-06-16 | 2020-06-16 | Procédé de génération d’images d’une caméra intérieure de véhicule |
PCT/EP2021/065160 WO2021254805A1 (fr) | 2020-06-16 | 2021-06-07 | Procédé de génération d'images d'une caméra intérieure de véhicule |
Publications (1)
Publication Number | Publication Date |
---|---|
EP4165605A1 true EP4165605A1 (fr) | 2023-04-19 |
Family
ID=73138891
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP21730593.7A Pending EP4165605A1 (fr) | 2020-06-16 | 2021-06-07 | Procédé de génération d'images d'une caméra intérieure de véhicule |
Country Status (5)
Country | Link |
---|---|
US (1) | US20230230359A1 (fr) |
EP (1) | EP4165605A1 (fr) |
CN (1) | CN115769267A (fr) |
FR (1) | FR3111460B1 (fr) |
WO (1) | WO2021254805A1 (fr) |
Family Cites Families (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5161204A (en) * | 1990-06-04 | 1992-11-03 | Neuristics, Inc. | Apparatus for generating a feature matrix based on normalized out-class and in-class variation matrices |
US8422797B2 (en) * | 2009-07-01 | 2013-04-16 | Honda Motor Co., Ltd. | Object recognition with 3D models |
US9437011B2 (en) * | 2012-06-11 | 2016-09-06 | Samsung Electronics Co., Ltd. | Method and apparatus for estimating a pose of a head for a person |
US9665800B1 (en) * | 2012-10-21 | 2017-05-30 | Google Inc. | Rendering virtual views of three-dimensional (3D) objects |
US9418467B2 (en) * | 2012-12-21 | 2016-08-16 | Honda Motor Co., Ltd. | 3D human models applied to pedestrian pose classification |
WO2016161553A1 (fr) | 2015-04-07 | 2016-10-13 | Intel Corporation | Génération et animations d'avatars |
US10068385B2 (en) * | 2015-12-15 | 2018-09-04 | Intel Corporation | Generation of synthetic 3-dimensional object images for recognition systems |
EP3427187A1 (fr) * | 2016-03-11 | 2019-01-16 | Siemens Mobility GmbH | Exploration de caractéristiques basée sur l'apprentissage profond pour une recherche d'image de détection 2,5d |
US9760827B1 (en) * | 2016-07-22 | 2017-09-12 | Alpine Electronics of Silicon Valley, Inc. | Neural network applications in resource constrained environments |
US10163003B2 (en) * | 2016-12-28 | 2018-12-25 | Adobe Systems Incorporated | Recognizing combinations of body shape, pose, and clothing in three-dimensional input images |
DE112017007252T5 (de) * | 2017-03-14 | 2019-12-19 | Omron Corporation | Fahrerüberwachungsvorrichtung, fahrerüberwachungsverfahren, lernvorrichtung und lernverfahren |
US10692000B2 (en) * | 2017-03-20 | 2020-06-23 | Sap Se | Training machine learning models |
US10235601B1 (en) * | 2017-09-07 | 2019-03-19 | 7D Labs, Inc. | Method for image analysis |
US10867214B2 (en) * | 2018-02-14 | 2020-12-15 | Nvidia Corporation | Generation of synthetic images for training a neural network model |
US10776642B2 (en) * | 2019-01-25 | 2020-09-15 | Toyota Research Institute, Inc. | Sampling training data for in-cabin human detection from raw video |
WO2020231401A1 (fr) * | 2019-05-13 | 2020-11-19 | Huawei Technologies Co., Ltd. | Réseau neuronal pour une estimation de pose de tête et de regard utilisant des données synthétiques photoréalistes |
US11487968B2 (en) * | 2019-12-16 | 2022-11-01 | Nvidia Corporation | Neural network based facial analysis using facial landmarks and associated confidence values |
US20210390767A1 (en) * | 2020-06-11 | 2021-12-16 | Microsoft Technology Licensing, Llc | Computing images of head mounted display wearer |
-
2020
- 2020-06-16 FR FR2006233A patent/FR3111460B1/fr active Active
-
2021
- 2021-06-07 CN CN202180042748.6A patent/CN115769267A/zh active Pending
- 2021-06-07 US US18/008,548 patent/US20230230359A1/en active Pending
- 2021-06-07 WO PCT/EP2021/065160 patent/WO2021254805A1/fr unknown
- 2021-06-07 EP EP21730593.7A patent/EP4165605A1/fr active Pending
Also Published As
Publication number | Publication date |
---|---|
FR3111460A1 (fr) | 2021-12-17 |
WO2021254805A1 (fr) | 2021-12-23 |
CN115769267A (zh) | 2023-03-07 |
FR3111460B1 (fr) | 2023-03-31 |
US20230230359A1 (en) | 2023-07-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2011015928A2 (fr) | Procede de traitement d'image pour corriger une image cible en fonction d'une image de reference et dispositif de traitement d'image correspondant | |
CN111612891A (zh) | 模型生成方法、点云数据处理方法、装置、设备及介质 | |
EP4046129B1 (fr) | Procédé de fabrication | |
CN115131492A (zh) | 目标对象的重光照方法、装置及存储介质和背景替换方法 | |
EP3614306A1 (fr) | Procédé de localisation et d'identification de visage et de détermination de pose, à partir d'une vue tridimensionnelle | |
EP3582141A1 (fr) | Procédé d'apprentissage de paramètres d'un réseau de neurones à convolution | |
EP4165605A1 (fr) | Procédé de génération d'images d'une caméra intérieure de véhicule | |
WO2019129985A1 (fr) | Procede de formation d'un reseau de neurones pour la reconnaissance d'une sequence de caracteres et procede de reconnaissance associe | |
DE112022002137T5 (de) | Hybrides differenzierbares Rendering für Lichttransportsimulationssysteme und -Anwendungen | |
EP4042319A1 (fr) | Procédé de reconnaissance d'objets à représentativité augmentée | |
EP3929809A1 (fr) | Procédé de détection d'au moins un trait biométrique visible sur une image d entrée au moyen d'un réseau de neurones à convolution | |
WO2011089517A1 (fr) | Procede et dispositif de reconnaissance de visages en conditions de poses variables | |
WO2021116635A1 (fr) | Procede et systeme de generation de modeles numeriques 3d | |
FR3059806B1 (fr) | Procede d'obtention d'un systeme de labellisation d'images, programme d'ordinateur et dispositif correspondants, systeme de labellisation d'images | |
CA3161385A1 (fr) | Procede de segmentation d'une image d'entree representant un document comportant des informations structurees | |
CA2439867A1 (fr) | Procede et dispositif de reperage d'un objet par sa forme, sa dimension et/ou son orientation | |
US20230055204A1 (en) | Generating colorized digital images utilizing a re-colorization neural network with local hints | |
FR3126253A1 (fr) | Procédé pour normaliser la variabilité d’une image, application de ce procédé à la détection d’anomalie et système d’inspection visuelle implémentant cette détection | |
WO2022171960A1 (fr) | Procédé d'apprentissage d'un système d'apprentissage automatique pour la détection et la modélisation d'un objet dans une image, produit programme d'ordinateur et dispositif correspondant | |
Pisarek | Manipulation of cognitive image properties | |
WO2023161531A1 (fr) | Système et procédé de segmentation d'image semi-automatique par apprentissage à boucle d'interaction utilisateur et procédé d'entraînement associé | |
FR3128015A1 (fr) | Système d’aide à la navigation d’un véhicule. | |
EP4207077A1 (fr) | Correction d'artefacts de reconstructions tomographiques par reseaux de neurones | |
Schmitt | Content-based image enhancement for eye fundus images visualization | |
Yoon et al. | FRAG: Frequency Adapting Group for Diffusion Video Editing |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: UNKNOWN |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
17P | Request for examination filed |
Effective date: 20230116 |
|
AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
DAV | Request for validation of the european patent (deleted) | ||
DAX | Request for extension of the european patent (deleted) | ||
RAP3 | Party data changed (applicant data changed or rights of an application transferred) |
Owner name: CONTINENTAL AUTOMOTIVE TECHNOLOGIES GMBH |