METHOD FOR AUTOMATICALLY GENErating an Ordered Image Sequence for Applying Differentiated Image Processing

Abstract

METHOD FOR AUTOMATICALLY GENERATING AN ORDERED IMAGE SEQUENCE FOR THE APPLICATION OF DIFFERENTIATED IMAGE PROCESSING. A computer-implemented method for selecting and ordering a set of images for image processing of a sequence of images, said method comprising: First classification (CLASS1) of each image in the first set (ENS1) according to context classes (CLc); Extraction of a first specification (SPEC1) given as a function of at least one first identified context class, said first specification (SPEC1) comprising an expected number of images, a plurality of given views of each expected image and an order (ORD1) of said expected images; Second classification (CLASS2) according to view classes (CLv) of each image in the first subset;Selection of a series of images (S1) from the first subset of images containing a given number of images, each selected image corresponding to a given view; Ordering of said series of images. Figure for the abbreviation: Fig. 1;

Description

Title of the invention: METHOD FOR AUTOMATICALLY GENERATING AN ORDERED SEQUENCE OF IMAGES FOR THE APPLICATION OF DIF IMAGE PROCESSING FERENCIES Field of invention

[0001] The field of the invention relates to methods and systems for automatically processing a large number of images for publication on a server. The field of the invention relates to methods and systems for automatically identifying and selecting images and for applying image processing to a given sequence of images. State of the art

[0002] Currently, there are methods for processing images according to a result to be achieved for their publication. These treatments may include cropping the image, masking areas to be removed or even extracting areas of interest.

[0003] However, a problem arises when it is necessary to process a large quantity of received images for their automatic publication. Generally, in this type of application of automated processing of a large number of images, it is necessary to process sequences of images to meet different publication wishes or constraints. For example, it is necessary to homogenize a rendering of a plurality of images, to apply strategies for representations and proportions of certain images. Furthermore, the subject of the image may also require specific processing, for example with regard to the anonymization of individuals represented in a photo or the exploitation of image data such as a label to be enlarged or an orientation to be standardized for an object. It is therefore necessary to consider each image or groups of images to apply specific processing to them.The need to apply differentiated processing to a group of images complicates the implementation of image processing to a plurality of images.

[0004] A problem with existing methods addressing these image processing operations is that there are generally processing steps that still need to be carried out "by hand", particularly because the processing operations to be applied to the images are heterogeneous, varied and often depend on the nature of the subject or object to be represented.

[0005] The implementation of automated processing can lead to numerous errors in the made of a large amount of data to be processed. In particular, these errors can occur due to poor detection of an image class, poor image quality or even a lack of information necessary to execute an automated process.

[0006] Furthermore, current automatic classification methods are too costly in terms of calculation time and consume computing resources,

[0007] There is a need to remedy the aforementioned drawbacks. Summary of the invention

[0008] According to a first aspect, the invention relates to a computer-implemented method for selecting and ordering a set of images for image processing of an image sequence, said method comprising: • Receiving a plurality of images defining a first set of images • First classification of each image of the first set according to context classes from a first learning function implementing a machine learning model trained from training data; • Extraction of a first given specification based on at least one first identified context class, said first specification comprising an expected number of images, a plurality of given views of each expected image and an order of said expected images; • Selection of a first subset of images from the first identified context class; • Second classification according to view classes of each image of the first subset selected according to the first given context class; • Automatically selecting a series of images from the first subset of images comprising a given number of images corresponding to the expected number of the extracted specification, each selected image corresponding to a given view of the plurality of given views of the extracted specification; • Ordering said series of images according to the first extracted specification and the images selected from the first series of images; • Renaming the selected images of the first series and saving said images with a first set of metadata comprising at least the context class for applying image processing according to a second data specification relating to processing to be applied to a sequence of images.

[0009] One advantage is that it allows for the generation of a sequence of ordered images to perform individualized processing on images for their subsequent publication.

[0010] According to one embodiment, the images received from the first set of images comprise pre-processing aimed at identifying duplicates of images from said first set and keeping only one image from a plurality of substantially identical images.

[0011] According to one embodiment, the images received from the first set of images comprise pre-processing aimed at verifying a minimum dimension of each image in order to authorize the cropping of an image when its dimensions are sufficient.

[0012] One advantage is to keep the images of better quality in order to generate the image sequence.

[0013] According to one embodiment, the images received from the first set of images are classified by the first classification, each image being associated with a score relating to the probability of belonging to said context class, a step of selecting a number of images having the best scores in each context class making it possible to select the images retained for the application of the second classification.

[0014] One advantage is to determine the context of the received images in order to determine the specification to select to apply an appropriate view selection.

[0015] According to one embodiment, a calculation of a score relating to the probability of belonging to said view class is carried out for each image classified by the first classification or by each image of the first set, a step of selecting a number of images having the best scores in each view class making it possible to select the images retained to generate the first sequence.

[0016] One advantage is to retain the most relevant views to be retained to generate an image sequence according to a selected specification.

[0017] According to one embodiment, the selection of the first subset of images extracted from the first set of images is determined based on a distribution obtained from the classification of a proportion of images classified in each context class.

[0018] According to one embodiment, the first selected specification is determined based on a distribution obtained from the classification of a proportion of images classified in each context class.

[0019] One advantage is to use all available information to detect the context class, namely: the number of images received, their distribution, the context classes detected, etc. One advantage is to make fewer errors in selecting a context class.

[0020] According to one embodiment, the received images of the first set of images include a first identifier relating to an entity. One advantage is to apply specifications specific to certain users in order to reduce specification errors detected when the algorithm includes many specifications.

[0021] According to one embodiment, the extraction of the first specification is carried out from a selection of specifications among specifications relating to the first identifier.

[0022] According to one embodiment, the context class may relate to: • a first context class corresponding to images comprising a subject such as an individual within the image, or; • a second context class corresponding to images containing a subject such as an object within the image, or; • a third context class corresponding to images containing a subject such as an object within the image.

[0023] One advantage is to recognize a context class in which certain processing will be performed for each image of the sequence to be generated. Typically, certain processing will be differentiated depending on whether an object is carried by a model such as a mannequin in an environment or whether it will only be represented in an environment with possibly a human in the environment.

[0024] According to one embodiment, a check of the number of views is carried out following the second classification, said check making it possible to verify the existence in each first set of at least one image classified according to each view class of the first specification.

[0025] According to one embodiment, when the control of the number of views does not comply with the value of the first specification, then an electronic notification is automatically sent to a remote server, said notification comprising an indicator of the missing image according to a view class.

[0026] One advantage is to ensure the correct number of views or to issue a notification with the aim of obtaining the missing views.

[0027] According to one embodiment, a product class of at least one image of the first set is determined by applying an image classification algorithm to classify input images according to a plurality of product classes.

[0028] One advantage is to optimize the detection of the specification which may depend in certain cases on the product class.

[0029] According to one embodiment, the image classification algorithm for classifying said image according to a product class is carried out: • By implementing a machine learning algorithm trained from a set of labeled training images and / or; • By implementing a machine learning algorithm trained from

[0030]

[0031]

[0032]

[0033]

[0034]

[0035] of a set of labeled file names. One advantage is to take advantage of maximum information to classify product classes. In an exemplary embodiment, when the file name or file name structure is decoded, a query is automatically generated to query a database to extract a product class. The data can then be aggregated with the image as metadata. According to one embodiment, the first specification is determined based on determining the product class of at least one image of the first set. According to one embodiment, the view class may relate to: • a first view class corresponding to images containing a close-up object within the image, or; • a second view class corresponding to images containing an object or subject present in a landscape within the image, or; • a third context class corresponding to images containing an object having a label within the image, • a fourth context class corresponding to images containing an object of a given type having a given first orientation within the image. According to a second aspect, the invention relates to a method for generating a second sequence of images comprising: • receiving a sequence of ordered images, said sequence being associated with a context class, each image of said sequence representing at least one view of a product, each image comprising a file name; • identification of the view class of each image; • identification of a product class of an object present in at least one image of the first sequence; • extraction of a second given specification as a function of the context class, the view class and the product class of at least one image of the sequence, said second specification comprising for at least one image of the sequence processing data to be applied to said image; • Image processing to generate a modified image; • Generation of a second sequence of images for publication on a WEB page containing said modified image. An advantage of this second method is that it allows an ordered sequence of images to be processed to automatically apply differentiated treatments to said images. The ordering, when compared to a processing specification, can allow the images in the sequence to be selected to apply this or that treatment.

[0036] This second method can be implemented, for example, following the implementation of a method for selecting and ordering a set of images for image processing of a sequence of images. Consequently, this second method can be implemented with a method of one of the embodiments of the invention. An advantage of implementing these two methods is being able to process them independently, for example, on two servers or jointly on the same server.

[0037] According to a second aspect, the method comprises decoding the file name of at least one image of the first received sequence, said decoding implementing a first component segmenting a set of blocks of alphanumeric characters composing said file name and implementing a machine learning algorithm to classify said file name among predefined classes.

[0038] An advantage of decoding the file name at this stage is to coordinate with the renaming algorithm of the first method for selecting and ordering a set of images for image processing of an image sequence.

[0039] According to one embodiment, the product class is obtained: • Either by applying an image processing algorithm to recognize a product class by a machine learning algorithm trained from a set of training images; • Either by decoding metadata transmitted with the first sequence of images; • Either by decoding a file name of an image from the first sequence in which the product class is encoded.

[0040] According to one embodiment, the processing data extracted from the second specification comprise: • Data defining a dimension and a product template to generate a cropping of an area delimiting said object and a positioning of said cropped area accordingly; • Data defining a horizon line of the image aimed at positioning a point of the object on said horizon line, the point being able to be for example the barycenter or the middle of the zone delimiting said object; • Cropping data of a label detected in an area of ​​said object of the image to extract said label if necessary to generate an image of the label according to a predefined dimension; • data for resizing the image, resolving the image or repositioning a characteristic point of the image in a new frame of the image; • object orientation data to apply a change in the orientation of the product in the image to generate a new image; • a predefined configuration data.

[0041] According to another aspect, the system for automated processing of a set of images for their publication on a digital medium comprises a first server making it possible to carry out the steps of the method for the selection and ordering of a set of images for image processing of a sequence of images and a second server making it possible to carry out the steps of the method for generating a sequence of images ready for their publication, said processed images being organized in the form of a sequence of ordered images. Brief description of the figures

[0042] Other characteristics and advantages of the invention will emerge on reading the detailed description which follows, with reference to the appended figures, which illustrate:

[0043] [Fig. 1]: a first example of steps of an embodiment of the method of the invention making it possible to generate a sequence of ordered images;

[0044] [Fig.2]: an example of steps of an embodiment of the method of the invention making it possible to apply image processing to a sequence of images generated by the method of the invention;

[0045] [Fig.3]: an example of system architecture of an embodiment of the invention.

[0046] [Fig.4]: a first example of an image of an individual wearing clothes among a set of received images;

[0047] [Fig.5]: a first example of processing the image of [Fig.4] for its pu application;

[0048] [Fig.6]: a second example of processing the image of [Fig.4] for its pu application;

[0049] [Fig.7]: a second example of an image of an individual in a situation with an object among a set of received images,

[0050] [Fig.8]: an example of automatic processing of the image of [Fig.7] for its pu application.

[0051] [Fig. 1] represents an embodiment of the invention and in particular examples of steps implemented to generate a sequence of images to which image processing will be applied.

[0052] The method comprises a first step, denoted RECi, of receiving a set of images ENSp. These images are preferably received on a server accessible from a data network. The server is denoted SERVi in [Fig.3] and the data network is denoted NETb; it may be, for example, the Internet network. However, according to another configuration, the invention applies to a private network, such as a network corporate or organizational network. The server then includes a communication interface for receiving digital data sent by another entity connected to the NETb data network. A PC, a server or a tablet or even a Smartphone can define the equipment from which the images are sent to the SERVi server.

[0053] An authentication server, such as the SERV2 server shown in [Fig. 3] can be implemented in order to obtain access to the SERVi server. The SERVi server can be configured to carry out data storage and therefore image storage and it can be configured to carry out the first processing operations, in particular those corresponding to the first steps of the method of the invention shown in [Fig. 1].

[0054] According to one embodiment, before implementing the first classification step CLASSi, a pre-processing step (not shown) can be implemented in order to delete images from the first set in order to carry out a preliminary selection of a subset of images. These pre-processing steps can correspond to a file name check of each image in order to verify a formatting of the name of each image. According to another example, a pre-processing step can correspond to a check of a size of an image, its dimensions or its resolution. An advantage of these first pre-processing steps is to carry out simple and inexpensive steps in terms of computation time initially to avoid implementing more significant calculations at a later stage, such as that of the application of a neural network or a statistical function to classify the images.

[0055] According to one example, a scan of the image data makes it possible to verify that the dimensions of an image are greater than a predefined threshold. If the dimensions of the images are greater than a given threshold, the image is retained, otherwise the image is not retained.

[0056] According to an exemplary embodiment, when there is only one image of a given class, the image is retained.

[0057] According to one example, when there are at least two images in a given class, the image with the largest dimension is kept.

[0058] According to an example, when there is only one or more images in a given class whose dimensions of all the images are less than a given threshold, a notification is issued. The notification is issued for example to a third-party server in order to inform that the re-issuance of a compliant image must be re-engaged. According to an exemplary embodiment, the generation of such a notification is automatically engaged by considering a decoded user identifier which is associated with an electronic mail address.

[0059] According to another example, when the dimensions are less than a given threshold and no other image of the image class with a better resolution can be selected, then a resolution increase step is initiated. This step is known in the Anglo-Saxon literature in the field of photography as an "upscale" step. This step aims to generate new pixels from an algorithm to "artificially" increase the resolution of the image. Artificial enlargement of the image is performed to overcome a critical threshold of image rejection.

[0060] This step can be carried out in conjunction with the sending of a notification, one advantage is to organize a sequence of images for their publication and to allow images to be published quickly while ensuring potential subsequent correction by replacing one image with another.

[0061] [Fig.l] represents a first step of classification CLASSi of the received images ENSp This first classification CLASSi of each image of the first set ENSi is configured to classify the images according to context classes CLc. CLASS i also designates the first classifier comprising the hardware and software means to implement the first classification.

[0062] According to one embodiment, the classification function is performed from a first learning function implementing a machine learning model trained from training data. The machine learning model may be, for example, a neural network such as a convolutional neural network, otherwise known as a convolutional neural network. Generally speaking, the first learning function may be any function comprising coefficients that are learned during a supervised training phase performed with images labeled according to the context classes CLc. According to a first example, the learning function may therefore group together matrices such as a convolutional neural network within which the coefficients are learned. According to a second example, the learning function may be a statistical function.

[0063] The first classification CLASSi aims to classify the images of the first set of images ENSi according to predefined classes. This first classification CLASSi can be associated with a score which is calculated in order to define a threshold beyond which the image is in a context class and below which the image is not in the context class. According to another example, a plurality of scores can be calculated for each context class for each image in order to discriminate the most likely context class. According to another example, when scores are close for the same image within several context classes, an additional calculation or control step can be implemented in order to carry out an additional automatic test or a verification by a human.

[0064] According to different embodiments, the score can take into account different criteria. A weighting of the criteria can be implemented in order to take into account one criterion more importantly than another criterion. Criteria include the statistic or probability of an image belonging to a class, the image resolution, the image dimensions, or any other criterion defining a property of the image.

[0065] The first classification CLASSi comprises at least two context classes CLc. According to one embodiment, the first classification comprises a first context class of an object called "worn", that is to say that the object is worn on an individual. This concerns for the most part objects of clothing, hat, bag or shoes. According to this embodiment, this first classification comprises a second context class called "overall photo" and better known in the Anglo-Saxon literature as a class called "Packshot". In this context class, an object is presented in a plain background. According to another example, the first classification comprises a third context class called "mood photo", which corresponds to a photo of an object in a setting or a landscape in which other objects may be present.

[0066] The identification of the object or the type of object can be carried out in different ways. A first way is to interpret data transmitted with the set of ENSi images. A second way is to decode in the file name corresponding to an image a reference linked to a product and to compare this reference with a database of known references. A third way is to execute a learning function implementing a machine learning model to identify a product class from the file name. This name can for example be segmented beforehand in the form of blocks of alphanumeric characters to decode each block, for example from a machine learning algorithm. A fourth way comprises the execution of an algorithm for detecting the product within the image for example by implementing a machine learning algorithm trained from a set of labeled images.

[0067] The product class CLp can advantageously be used to select a first specification SPECi for example in combination with other data such as the context class and / or a user identifier.

[0068] The step of classifying an image according to a plurality of product classes CLP is noted CLASS3. It is preferably carried out during the first phase of the method aiming to generate the first sequence SEQi, however this step can also be implemented in the second phase of the method aiming to generate the second sequence SEQ2 of images transformed from the images of the first SEQi. [Fig.2] represents this step in this second phase. One interest is to enrich the classification of the object of the image within a product class to adapt particular treatments to the image. Another interest is to corroborate data which can be received in entry identifying the product that was not verified in the first phase of the method.

[0069] When the context class CLc is identified using the first classification, the image is associated with this context class. According to one example, this association corresponds to a labeling of the image, that is to say the creation of a metadata which is recorded for example in a database in which an identifier of the image is also recorded. According to one example, the association comprises the recording of the images of a given context class CLc in the same directory. According to another example, a renaming of the image file is carried out.

[0070] The set of images of the set of received images ENSi is therefore classified according to the first classification CLASSi. According to one embodiment, a rule is defined according to the distribution of the classes in which the images have been classified to extract a first given specification SPECi from a memory. The first given specification SPECi comprises parameters and a set of processing operations or actions to be carried out on the set of images ENSi or a part of this set ESNi in order to generate an ordered sequence of images from the first set ENSi.

[0071] According to one embodiment, a given first SPECi specification is associated with a given context class.

[0072] In this case, different embodiments of the invention make it possible to define different rules making it possible to select the first SPECi specification as a function of one or more context classes CLc identified by the set of ENSi images. According to a first example, when at least one context class corresponds to the “scope” class, then a given first SPECi specification is automatically selected.

[0073] According to a second example, the majority proportion of the most representative context class makes it possible to identify a given CLc specification class.

[0074] According to a third example, a given number of images in the “overall photo” class makes it possible to select a given first CLc specification.

[0075] According to a fourth example, the presence of at least one image in the “overall photo” class and at least one image in the “ambient photo” class makes it possible to select a first given specification.

[0076] According to another example, the first SPECi specification is determined from data associated with the set of received images. This may be metadata recorded in a configuration file, or metadata encoded in the image or even data encoded in the file name of at least one image of the first ENSi set.

[0077] The first specification SPECi is therefore extracted automatically following the reception of a set of ENSi images to be processed. This step of extraction EXTi of the first specification SPECi is noted EXTi in [Fig.l]. The first speci SPECi fication includes an expected number of images noted Nbb a plurality of given views Vli of each expected image and an order ORDi of said expected images.

[0078] For example, a first SPECi specification comprises the definition of a series of images corresponding to a series of 4 images, such as a first front view IMi of an image of a carried context class, a second back view IM2 of an image of a carried context class, a third side view IM3 of an image of a carried context class and a fourth front view IM4 of an image of an overall photo context class.

[0079] According to another example, a first SPECi specification comprises the definition of a series of images corresponding to a series of two images represented schematically in Figures 7 and 8 of an object, respectively represented here by a chainsaw, according to a first view of an image of an ambient image context class in [Fig.7] and according to a second view of an image of an overall photo context class in [Fig.8].

[0080] According to another example, this last series could have been completed with other photos for example through another first SPECi specification which can include a series of 3 or 4 images for example.

[0081] Accordingly, the first SPECi specification includes a number of expected images and a number of image views according to given context classes.

[0082] Furthermore, the first SPECi specification includes an order, or even an ordering, of the expected images. In the first case, IMb IM2, IM3 and IM4 can be ordered by their index and thus be numbered in the order of their sequence. One advantage is to allow dedicated processing to be carried out on each of the images in a later step for their automatic publication.

[0083] In order to identify the different views among the different images of the first set ENSi classified according to context classes CLc, a second classifier CLASS2 of views is implemented. The designation CLASS2 designates both the second classifier and the second classification step. The second classifier comprising the hardware and software means for implementing the second classification.

[0084] The second classification CLASS2 comprises the classification of all or part of the images classified by the first classification. According to one example, the second classification will be implemented only on images classified according to a class of the first classifier CLASSi. According to another example, the second classification CLASS2 is applied only to images having a file name according to a certain predefined formalism.

[0085] According to one embodiment, the classification function CLASS2 is carried out from a second learning function implementing a learning model machine trained from training data. The machine learning model can be, for example, a neural network such as a convolutional neural network, otherwise known as a convolutional neural network. Generally speaking, the second learning function can be any function comprising coefficients that are learned during a supervised training phase carried out with images labeled according to the CLv view classes. In a first example, the learning function can therefore group together matrices such as a convolutional neural network within which the coefficients are learned. In a second example, the learning function can be a statistical function.

[0086] Each context class includes a set of possible views of the object. Training is therefore preferentially carried out for a given context class.

[0087] The second classification CLASS2 aims to classify the images of a given context class of images of the first set of images ENSp. This second classification CLASS2 can be associated with a score which is calculated in order to define a threshold beyond which the image is in a view class CLv and below which the image is not in the view class CLv. According to another example, a plurality of scores can be calculated for each view class CLv for each image in order to discriminate the most likely view class CLv. According to another example, when scores are close for the same image within several view classes CLv, an additional calculation or control step can be implemented in order to carry out an additional automatic test or a verification by a human in order to assign the correct view class Civ. The intervention of a human can in particular be planned in a learning phase in order to enrich the images with labels.The view class can also be corrected automatically after receiving a notification indicating a bad classification of the image. The notification can take the form of a report in a digital format for example .doc, .pdf issued through the sending and receiving of a message.

[0088] According to different embodiments, the score can take into account different criteria. A weighting of the criteria can be implemented in order to take into account one criterion more importantly than another criterion. Among the criteria are the statistics or the probability of an image belonging to a class, the resolution of the image, the dimensions of the image or any other criterion defining a property of the image.

[0089] The second classification CLASS2 comprises at least two view classes CLv. According to one embodiment, the second classification CLASS2 comprises a first view class CLvi of an object called “close-up”, that is to say that the object is in the foreground of the image and occupies a certain proportion of the image. It can be any object represented in a photo, such as a piece of furniture, an item of clothing, a consumer good, a tool, etc. According to this embodiment, this second CLASS2 classification includes a second CLv2 class relating to images representing an object in a wide shot.

[0090] A third CLv3 view class, called "right orientation" and a fourth CLv4 class called "left orientation" can also be defined. These classes make it possible to identify an orientation of an object such as shoes or an individual's head or their gaze.

[0091] In this example, the third class CLv3 and the fourth class CLv4 can be complementary to the first class and second class, that is, they are not exclusive between these groups of classes. However, when the third class is associated with an image, the fourth class cannot be. In this case, these classes are exclusive of each other.

[0092] The invention makes it possible to configure rules for exclusion or complementarity of view classes of a context class.

[0093] According to another example, a fifth CLv5 class makes it possible to classify configurations of an object. The configurations correspond to variations of representation of an object. For example, a first configuration corresponding to a CLv5i subclass is an open configuration, a second configuration corresponding to a CLv52 subclass is a closed configuration, a third configuration corresponding to a CLv53 subclass is an exploded configuration, that is to say in which all the parts of an object are represented. In the case of a piece of furniture, such as a chest of drawers having two doors and a drawer, the first configuration corresponds to the closed chest of drawers, the second configuration corresponds to the chest of drawers having an open door and the third configuration corresponds to the chest of drawers in which the drawer has been fully or partially extracted and the two doors are open.

[0094] It is understood that an image of the first set having a context class CLc2 can have the following combination of classes {CLvi, CLv3, CLv5i}.

[0095] According to another example, a sixth CLv6 class and a seventh CLv7 class may be classes making it possible to classify images representing an object having a visible or invisible label. One advantage of this classification is to generate a specific processing aimed, for example, at enlarging a portion of the image comprising a label in the area of ​​the image in which there is a label.

[0096] According to one example, when a classification results in a class error or a class inconsistency, for example by associating mutually incompatible classes with an image, a NOTIFi notification may be issued automatically so as to require an automatic or human verification of a class associated with an image. The notification may take the form, for example, of a report published in a format digital for example .doc, .pdf.

[0097] When all the images of the selection of images classified by the second classification CLASS2 have been processed, a second selection of images SEL2 can be implemented to choose the number of images corresponding to the number of images predefined in the first specification SPECi.

[0098] According to an example, a SPECi specification comprises a list of 4 images having 4 different view classes and a given order of these views. According to this same example, a set of 10 images is received in the set ENS1. First pre-processings make it possible to discard three images from the list of 10 images. A first discarded image is considered to be blurred, a second image is considered to be a duplicate of another image and a third image corresponds to an image error, for example because the object represented in this image is different from all the other objects in all the other images.

[0099] In this example, there remain 7 images to be classified using the second CLASS2 classification since 3 images were discarded by the pre-processing. Among the classified images, 4 images are retained because the SPECi specification includes a number of 4 images.

[0100] In this example, among the 7 classified images, 2 images are classified in the same first expected view class, 2 other images are classified in a second expected view class, another image is classified in a third expected view class, another image is classified in a fourth expected view class, finally a last image is classified in an unexpected view class. An expected view class is understood to mean a view class specified in the first specification SPEC i and an unexpected view class is understood to mean a view class not specified in the specification SPEC p

[0101] In order to determine whether two images are duplicates of each other, an algorithm may be implemented to compare data between the two images under consideration. The data compared may be metadata and / or pixel data defining the image itself.

[0102] Metadata can be the file name, file size, or any other data encoded in the file.

[0103] According to an example that can be combined with the algorithm that performs the comparison of the metadata, another algorithm compares groups of pixels of each image. Indeed, a duplicate can be determined by comparing 10 points each defining one or more pixels of each image. The compared pixels or groups of pixels defining points can be multiplied so as to compare different groups of pixels.

[0104] According to one example, the data set of two images is compared between them. For example, a percentage of similar pixels between the two images can be calculated. In this case, a threshold percentage can be set to determine if two images are duplicates of each other.

[0105] According to an exemplary embodiment, a first comparison algorithm comparing a few points between two images is implemented. In the case of a similarity of the comparison points, a second algorithm is applied which makes it possible to compare a larger number of pixels between the two images. The implementation of such a first comparison algorithm makes it possible to reduce the calculation times when a large number of images must be processed while ensuring a fine comparison by a second comparison algorithm when duplicate image indices are noted.

[0106] According to another scenario, two images comprising the same class of a classifier can be annotated as potential duplicates of each other. Typically, when the algorithm detects a characteristic element of an image, such as an object, for example an umbrella, in each image, an algorithm configured to detect objects in an image can be implemented to generate a “duplicate to be confirmed” indicator. In a second step, a more refined algorithm makes it possible to verify whether the images are actually duplicates of each other.

[0107] The method of the invention makes it possible to implement a score calculation step making it possible to automatically assess the best classification between two images having the same class at the end of the second classification CLASS2. This score can correspond to a statistical calculation of likelihood or even a calculation of distance according to a metric defined by the classifier.

[0108] In the latter case, the 4 images corresponding to the images with the best scores in each expected view class can be selected. This selection step is noted SEL2 in [Fig.l].

[0109] According to one embodiment, in a second step, when the images are selected for each of the views, the method makes it possible to order them according to the predefined order which is defined in the first selected specification. The method therefore makes it possible to generate a first sequence of images SEQi.

[0110] In one embodiment, the method of the invention comprises a step for renaming the selected and ordered images.

[0111] A first sequence SEQi may have a given type or identifier allowing it to be processed in a certain way. For example, a sequence of such a type may be sent to a pre-processing server to apply intermediate processing, or it may be sent for quality control because the sequences in question must be labeled in a learning phase or they are sent to a server to apply particular image processing.

[0112] According to one embodiment, a common processing is applied to all of the images of the selection, these treatments may include resizing to generate images of the same dimensions or a treatment aimed at generating images of the same resolutions, either by reducing the resolution of certain images or by increasing the resolution of certain images. According to another example, a treatment applied to each image aims to embed the same graphic element in the image preferentially at the same position in the image. This may be a logo, a specific mention, or an indication of a copyright, that is to say the presence of a copyright within the image. According to another example, it may be a background common to each image. Example of a first SPEC1 specification

[0113] According to an example, a first SPECi specification comprises for a given object and a given identifier corresponding to a given user a list of desired views of the object, each view defining an image. This list of desired views comprises expected characteristics. If the object or the product is a pair of shoes, the first SPECi specification may comprise for a given view a right orientation of the shoes, a place of the shoes in the image expressed in proportion to the image. In addition, the first specification may comprise an ordered list of images of different context classes. For example, the first SPECI specification may comprise a list of four images of context classes CLc of the “overall photo” type and a context class CLc of the “scope” type.

[0114] The invention makes it possible to chain two treatments specific to a group of images so as to optimize calculation times and resources. The first processing chain aims to organize an ordered sequence of images and the second processing chain aims to process each image in a differentiated manner. To this end, the first SPECI specification includes information making it possible to recognize images from a set of images to prepare the treatments to be carried out on said images. The data making it possible to parameterize the treatments are specified in a second SPEC2 specification.

[0115] However, depending on the embodiments, the invention makes it possible to carry out pre-processing during the development of the first sequence SEQi and conversely, detections can be carried out or enriched or adjusted during the development of the second sequence SEQ 2. Thus, the method of the invention can be configured on a case-by-case basis by enriching or lightening this or that first specification SPCi or second specification SPEC2.

[0116] The second specification SPEC2 which will be applied later is considered for at least one image of the sequence SEQi in order to apply particular image processing for a given image of the first sequence.

[0117] The invention covers the following two cases: • A single second specification SPEC2 is extracted and includes all processing to be applied to each image of a first SEQi sequence, for this purpose an identification of a type of first SEQi sequence makes it possible to identify the correct second SPEC2 specification to be extracted; • A plurality of second SPEC2 specifications are extracted and each of them comprises a processing to be applied to an image or a group of images of the first sequence SEQi. Transmission of the sequence SEQ1

[0118] According to one embodiment, the image sequence SEQi is transmitted to another remote server SERV3 in order to carry out other processing for the automatic publication of the images. According to another embodiment, the sequence SEQi is processed on the same server SERVi having carried out the processing to generate this first sequence SEQb.

[0119] The remainder of the description is described according to the mode implementing another SERV3 server, however the invention applies within the framework of an implementation of all the steps carried out on the same SERVi server.

[0120] The server SERV3 therefore receives a new sequence SEQi of images to be processed. According to one embodiment, the server SERV3 also receives operating data from the server SERVi. This operating data may correspond, for example, to data describing characteristics of the images in the sequence of images SEQi, such as the resolution, dimensions, view types, framing, or metadata such as context classes or view classes or data relating to identification characteristics or position in the image of certain elements, such as the identification of an object and its position in the image, or even an area in which it is located, etc.

[0121] According to an exemplary embodiment, the transmitted operating data corresponds to the first SPECi specification.

[0122] Image processing to generate the second sequence SEQ2

[0123] According to one embodiment, a set of machine learning algorithms are implemented according to: • data from the first SPECi specification and / or; • of the context class CLc associated with the first sequence SEQi and / or; • the name(s) of the image files and / or; • other operating data transmitted by the first SERVi server with the first SEQi sequence of images.

[0124] One interest is to apply a processing adapted to the first sequence of images received SEQb

[0125] To this end, the processing carried out by the SERV3 server makes it possible to detect certain data from the images, allowing the correct transformations and mo modifications of the images from the first SEQi sequence to generate a second SEQi sequence ready for publication.

[0126] Among the algorithms, the present invention proposes to implement a plurality of algorithms which can be more or less selected or combined with each other in order to carry out several processing operations on the same processing chain of a first sequence SEQb.

[0127] According to a first embodiment, a first ALGi algorithm makes it possible to detect the class of products present in the image.

[0128] According to a first example, this information is known upstream of the processing aimed at generating the second sequence SEQ2 for example because it was determined during the processing to generate the first sequence SEQi, or transmitted with the first sequence within a metadata such as a label or even when it is encoded in the name of the file of at least one image of the first sequence SEQb

[0129] According to a second example, this information is not known in advance.

[0130] The ALGi algorithm makes it possible to identify a class of CLp products and to verify that the known information, if applicable, was correct or to associate this class of CLp products with the sequence of images.

[0131] This product class CLp can be combined with each view class CLv to extract a second specification SPEC2 in order to generate a given treatment. The step of extracting the second specification is noted EXT2 in [Fig.2]. The treatments are represented in [Fig.2] by the notations TIb TIk, TiN.

[0132] For example, if the context class CLc is a given “overall photo”, the view class CLv of the image considered is a “close-up” type class and the product class CLp is a household appliance, such as a washing machine or a refrigerator, then the second extracted specification SPEC2 may include data aimed at removing the background of the image and cropping and recentering the product according to a predefined dimension and a predefined template. Finally, this second specification SPEC2 may include instructions for launching a second algorithm ALG2 for detecting the presence of a label on the product.

[0133] According to one example, the template comprises the dimensions of a polygon, such as a rectangle, positioned on a horizon line and positioned with respect to lateral margins and longitudinal margins. Example of label detection

[0134] The second algorithm ALG2 is then launched in order to detect a label on the product. If a label is present, the second specification further includes data for cropping this label and generating a new image of a close-up of the label. The dimensions and framing template(s) of the label can then be extracted from the second specification SPEC2.

[0135] According to an alternative, this algorithm was carried out during the steps aimed at generating a first sequence SEQi. The invention makes it possible to implement this step upstream or downstream of the transmission of the first sequence SEQi. When this step is carried out upstream of the transmission of the first sequence SEQi, the result of the detection is transmitted with the images of the first sequence SEQi in the form of metadata for example.

[0136] When no label is received then a NOTIF2 notification can be issued to check that the product class is the correct one or that the context class is the correct one. According to a first example, the NOTIF2 notification is issued to one of the steps of [Fig.l], for example, to a software component making it possible to re-challenge the detection of the context class or the view class. [Fig.2] represents such a NOTIF2 notification which is issued and [Fig.l] represents this same notification which is received by the block of the first classifier CLASSi. The NOTIF2 notification can take for example the form of a report in a digital format for example .doc, .pdf issued thanks to the sending and receiving of a message. Example: orientation detection

[0137] According to one embodiment, when a given product class CLp is obtained from a first algorithm ALGi for detecting or classifying the product class of the image and a given view class CLv associated with the same image of a first sequence SEQi, an algorithm for detecting the orientation of the product can be launched.

[0138] According to an alternative, this algorithm was carried out during the steps aimed at generating a first sequence SEQi. The invention makes it possible to implement this step upstream or downstream of the transmission of the first sequence SEQi. When this step is carried out upstream of the transmission of the first sequence SEQi, the result of the detection is transmitted with the images of the first sequence SEQi in the form of metadata for example.

[0139] According to an example, this may be the orientation of a shoe or any other accessory that may be presented in different orientations. The detection of an orientation may consist of detecting a right orientation or a left orientation, that is to say an orientation of the product towards a right edge or a left edge of the image.

[0140] The detection of the orientation of an object can be carried out in particular by using a depth map of the image and a machine learning algorithm taking into consideration a 3D image, or a 2D image and a depth map or even a 3D point cloud.

[0141] An advantage of this detection is to orient all objects in the same way either by choosing and applying a straight orientation, or by choosing a left orientation. This application can be achieved by generating a mirror image or maintaining the original orientation.

[0142] The generation of a mirror image with respect to an original image is obtained by reversing the image according to an axial symmetry.

[0143] The second SPEC2 specification may include orientation data so as to create a homogeneity effect in a visual rendering aimed at displaying several hundred images. The orientations thus modified make it possible to create an intuitive reference for the user consulting the proposed images.

[0144] According to another embodiment, a third algorithm ALG3 makes it possible to detect a configuration of a product such as data characteristic of its usage configuration such as “open”, closed”, “on”, “off” or even data characteristic of the product such as its size, its color, its model, etc.

[0145] For this purpose, a machine learning algorithm can be used to classify images automatically to define sets of images classified according to a characteristic of the product or more generally of the object. The learning can be supervised or unsupervised. A training data set generally makes it possible to improve the classification of objects. Typically, training with images comprising different models of the same type of product or object can make it possible to recognize a given model among different models. An algorithm for analyzing the colors of objects can be less constrained on the labeling of the training data.

[0146] One advantage of a classification is that it allows the generation of image effects such as ambient effects, semantic data or hyperlinks, for example, to generate two alternative views of a piece of equipment in the open and closed position. When they are published on a digital platform, the generation of links and data to facilitate navigation can be carried out automatically from the classes associated with the images. Typically, the classifications of an image within a class can lead to the automatic creation of a miniature image for its representation as a thumbnail or an interactive inscription inviting a user to perform an action on the image to switch to another view.

[0147] Example in which the context class allows the cropping to be adjusted

[0148] According to another embodiment, the detection of a context class CLc called “ambient photo” makes it possible to generate a first image sequence SEQi for which specific processing will be carried out as a function of said context class and as a function of the detected object. In particular, if the product class corresponds to a given tool, such as a chainsaw 20 shown in [Fig.7], within a landscape comprising trees 25, a log of wood 26 and an individual 1', then the processing of the image of [Fig.8] may be adapted as a function of the second SPEC2 specification to center the object and crop it according to a given template and dimension.

[0149] In the case where an image is classified in a context class of the “mood photo” type and it includes a garment worn by a mannequin, then the second specification SPEC2 can include data corresponding to another template and another dimension making it possible to generate a cropping and an adapted positioning. [Fig.8] represents such a cropping in which the chainsaw is represented on a base 22 and includes a label 21. In this example, an image background was used to generate a modified image within the second sequence seq2.

[0150] According to one embodiment, the first sequence SEQi is associated with a “worn” context class. This sequence of images may comprise a plurality of images including the image of [Fig.4]. Other images of the sequence SEQi may be used and are not shown, such as an image of the jacket and pants 10 without a mannequin or an image of the pants and jacket worn by the individual shown from behind.

[0151] In order to determine a second SPEC2 specification appropriate to the image processing to be applied, the CLv view class can be used for each image in order to extract the appropriate data to be used to carry out a suitable processing on a certain view of an object in the image.

[0152] When the method of the invention detects a product class of the “clothing” type, this product class associated with the “worn” context class and a “front and full-length” view class makes it possible to extract a second given specification. In this case, this second SPEC2 specification may include cropping data aimed at cutting the image at the level of the eyes 2 in order to respect the anonymity of the mannequin. This image is represented in [Fig. 5]. For this, the second SPEC2 specification may include the instructions for generating a fourth ALG4 algorithm for detecting the eyes 2 of an individual 1. Such an algorithm may implement a machine learning model such as a CNN-type convolutional network. When this fourth ALG4 algorithm is launched on this image, the positions of the eyes 2 make it possible to automatically generate a cutting axis 5.

[0153] According to another example, a processing aims to enlarge a portion of a cut-out image in a new image or in the same image.

[0154] Thus, the invention makes it possible to take advantage of a context classification CLc, a view classification CLv and a product class CLp in order to generate appropriate image processing.

[0155] Thus, a new image is generated corresponding to image 5 in which individual 1 is represented without the upper part of the head, the part beyond the cutting line passing through the eyes 5.

[0156] Still in this same example, a new image is created along another cutting line, line 6. This line can be automatically generated based on a dimension and a template present in the second specification. The dimension and the template can correspond to a percentage of a dimension extracted from the image. For this purpose, a shape recognition algorithm can be used. This image is shown in [Fig.6].

[0157] Example of a multi-detection of product / object at different stages

[0158] It is specified, according to the embodiments of the invention, that the detection of a product class CLp can be carried out during the steps aimed at generating a first sequence SEQi or it can be carried out during the steps aimed at generating a second sequence SEQ2. An enriched detection of a given product class CLp can also be configured during the steps aimed at generating the second sequence SEQ2 by collecting more detailed data on the product. This step can be more costly in terms of time and computing resources than a simple detection. This enriched product detection can be complementary to a first product detection aimed at identifying, more quickly in the steps of generating the first sequence SEQi, a product class so as to optimize the computing times when numerous sets of images must be processed.

[0159] Simple detection means detection aimed at roughly classifying a product into a broad product family and enriched detection means detection aimed at collecting several characteristics of the product.

[0160] Example of contour or shape extraction

[0161] According to an exemplary embodiment, the second specification SPEC2 may comprise, for a given view or for a plurality of views, data enabling the execution of a fifth algorithm ALG5 aimed at cropping the representation of an individual in an image in order to extract it from said image. The method thus makes it possible to generate a new image with a modification of the background of the image. For this purpose, an image processing and shape contour analysis algorithm may be used. This algorithm may also be combined with or substituted for a depth map processing algorithm to segment portions of the image which are not in the same image plane. Example of filtering applied to an image

[0162] According to one embodiment, according to the context class CLc associated with the first sequence SEQi and the identified product class CLp, a second specification SPEC2 is extracted and used so as to extract data relating to a filtering of the image. The filter applied may consist of an ambient filter, i.e. the application of a modification of the brightness, saturation and / or hue of the image. A sharpness filter or an increase or reduction of depth effects from the depth map can be applied. Example of refocusing an image

[0163] According to one embodiment, according to the context class CLc associated with the first sequence SEQi, the view class CLv and the identified product class CLp, a second specification SPEC2 is extracted and used so as to extract data relating to the refocusing of a portion of the image. The portion of the refocused image preferably corresponds to the portion of the image delimiting the object identified in the image. The second specification SPEC2 may comprise the definition of a margin to be respected when manipulating the area to be refocused in order to regenerate a new refocused image from the image of the first sequence SEQi. Adding images in the second sequence

[0164] According to one embodiment, the second sequence SEQ2 comprises as many images as the first sequence SEQi. According to another embodiment, the second sequence SEQ2 may comprise more images than the sequence SEQb. Finally, according to another embodiment, the sequence SEQ2 comprises fewer images than the sequence SEQb. The number of images of the second sequence SEQ2 may be extracted from the second specification SPEC2.

[0165] According to an exemplary embodiment, the second sequence SEQ2 may be ordered according to the same order as the first sequence SEQb. According to an exemplary embodiment, the second specification SPEC2 comprises an order different from the order of the first specification SPECi. According to an exemplary embodiment, a second specification SPEC 2 may comprise a re-indexing of the images of the first sequence SEQi and the insertion of a new image with a new index to generate a second sequence of modified images SEQ2.

[0166] According to one embodiment, an ordering and number of images instruction are reapplied to the second sequence SEQ2 in order to generate a third sequence SEQ3 which is adapted to a given publication platform.

[0167] When all the processing operations are applied to the images of the first sequence SEQ i, the correct number of images is present in the second sequence SEQ2 or the third sequence SEQ3 and the ordering of the images in the sequences complies with the second specification SPEC2, the second sequence of images SEQ2 or the third sequence of images SEQ3 is generated and recorded in a memory.

[0168] According to one embodiment, the second sequence SEQ2 or the third sequence of images SEQ3 is either processed by the server SERV3 for automatic publication to a WEB site or a third-party digital platform, or sent to another remote server to be subsequently used by a system for publishing images on a digital medium.

Claims

Claims

1. A computer-implemented method for selecting and ordering a set of images for image processing of an image sequence, said method comprising: • Receiving a plurality of images defining a first set of images (ENSi); • First classification (CLASSi) of each image of the first set (ENSi) according to context classes (CLc) from a first learning function implementing a machine learning model trained from training data; • Extraction of a first specification (SPECi) given as a function of at least one first identified context class (CLci), said first specification (SPEC i) comprising an expected number of images (Nbi), a plurality of given views (Vli) of each expected image and an order (ORDi) of said expected images; • Selection of a first subset of images (ENSi') of the first context class (CLci) identified; • Second classification (CLASS2) according to view classes (CLv) of each image of the first subset (ENSi') selected according to the first context class (CLc) given; • Automatic selection of a series of images (SJ of the first subset of images (ENSi') comprising a given number of images corresponding to the expected number (Nbi) of the extracted specification (SPECi), each selected image corresponding to a given view of the plurality of given views (Vli) of the extracted specification (SPECi); • Ordering (ORDi) of said series of images according to the first extracted specification (SPECi) and of the images selected in the first series of images (Si); • Renaming (RENi) of the selected images of the first series (Si) and recording of said images with a first set of metadata comprising at least the context class (CLc) for application of image processing according to a second data specification (SPEC2) relating to a treatment to be applied to a sequence of images.

2. Method according to claim 1 characterized in that the images received from the first set of images (ENSi) comprise a pre-processing aimed at identifying duplicates of images from said first set and keeping only one image from a plurality of substantially identical images.

3. Method according to claim 1 characterized in that the images received from the first set of images (ENSi) comprise a pre-processing aimed at verifying a minimum dimension of each image in order to authorize the cropping of an image when its dimensions are sufficient.

4. Method according to claim 1 characterized in that the images received from the first set of images (ENSi) are classified by the first classification (CLASSi), each image being associated with a score relating to the probability of belonging to said context class (CLc), a step of selecting a number of images having the best scores in each context class making it possible to select the images retained for the application of the second classification (CLASS2).

5. Method according to claim 1 characterized in that a calculation of a score relating to the probability of belonging to said view class (CLv) is carried out for each image classified by the first classification (CLASSI) or by each image of the first set (ENSI), a step of selecting a number of images having the best scores in each view class making it possible to select the images retained to generate the first sequence (SEQi).

6. Method according to claim 1 characterized in that the selection of the first subset of images (ENSi') extracted from the first set of images (ENSi) is determined as a function of a distribution obtained from the classification of a proportion of images classified in each context class (CLc).

7. Method according to claim 1 characterized in that the first specification (SPECi) selected is determined as a function of a distribution obtained from the classification of a proportion of images classified in each context class (CLc).

8. Method according to claim 1 characterized in that the images received from the first set of images (ENSi) comprise a first identifier relating to an entity (IDc), said extraction of the first specification (SPECi) being carried out from a selection of specifications ({SPECi}) among specifications relating to the first identifier (IDc).

9. Method according to claim 1 characterized in that the context class (CLc) can relate to: • a first context class (CLci) corresponding to images comprising a subject such as an individual within the image, or; • a second context class (CLc2) corresponding to images comprising a subject such as an object within the image, or; • a third context class (CLc3) corresponding to images comprising a subject such as an object within the image.

10. Method according to claim 1 characterized in that a check of the number of views is carried out following the second classification (CLASS2), said check making it possible to verify the existence in each first set (ENSi) of at least one image classified according to each view class of the first specification (SPECi).

11. Method according to claim 10 characterized in that when the control of the number of views does not comply with the value of the first specification, then an electronic notification (NOTIF1) is automatically sent to a remote server, said notification comprising an indicator of the missing image according to a view class.

12. Method according to claim 1 characterized in that a product class (CLp) of at least one image of the first set (ENSi) is determined by the application of an image classification algorithm to classify input images according to a plurality of product classes (CLp).

13. Method according to claim 12 characterized in that the image classification algorithm for classifying said image according to a product class (CLp) is carried out • By implementing a machine learning algorithm trained from a set of labeled training images and / or; • By implementing a machine learning algorithm trained from a set of labeled file names.

14. Method according to claim 12 characterized in that the first specification (SPECi) is determined as a function of the determination of the product class (CLp) of at least one image of the first set (ENSû.

15. Method according to claim 1 characterized in that the view class (CLv) can relate to: • a first view class (CLvi) corresponding to images comprising a close-up object within the image, or; • a second view class (CLv2) corresponding to images comprising an object or a subject present in a landscape within the image, or; • a third context class (CLv3) corresponding to images comprising an object having a labeling within the image, • a fourth context class (CLv4) corresponding to images comprising an object of a given type having a given first orientation within the image.

16. Method according to any one of claims 1 to 15 characterized in that it comprises: • reception of a sequence of ordered images (SEQi), said sequence being associated with a context class (CLc), each image (IMi) of said sequence (SEQi) representing at least one view of a product, each image (IMi) comprising a file name; • identification of the view class (CLv) of each image (IMi); • identification of a product class (CLp) of an object present in at least one image of the first sequence (SEQi); • extraction of a second given specification (SPEC2) as a function of the context class (CLc), the view class (CLv) and the product class (CLp) of at least one image of the sequence (SEQi), said second specification (SPEC2) comprising for at least one image of the sequence (SEQi) processing data to be applied to said image; • Image processing (TIb TIk, TIN) to generate a modified image; • Generation of a second sequence (SEQ2) of images for their publication on a WEB page containing said modified image.

17. Method according to claim 12 or 16 characterized in that it comprises the decoding of the file name of at least one image of the first sequence (SEQi) received, said decoding implementing a first component segmenting a set of blocks of alphanumeric characters composing said file name and implementing a machine learning algorithm to classify said file name among predefined classes.

18. Method according to claim 12 or 16 characterized in that the product class (CLp) is obtained: • Either by the application of an image processing algorithm to recognize a product class (CLp) by a machine learning algorithm trained from a set of training images; • Or by the decoding of metadata transmitted with the first sequence of images (SEQi); • Or by the decoding of a file name of an image of the first sequence (SEQ1) in which the product class (CLp) is encoded.

19. Method according to any one of claims 16 to 18 characterized in that the processing data extracted from the second specification (SPEC2) comprise: • Data defining a dimension and a product template to generate a cropping of an area delimiting said object and a positioning of said cropped area accordingly (SPEC2); • Data defining a horizon line of the image aimed at positioning a point of the object on said horizon line, the point being able to be for example the barycenter or the middle of the area delimiting said object; • Cropping data of a label detected in an area of ​​said object of the image to extract said label if necessary to generate an image of the label according to a predefined dimension; • data for resizing the image, resolving the image or repositioning a characteristic point of the image in a new frame of the image; • object orientation data to apply a change in the orientation of the product in the image to generate a new image; • a predefined configuration data.

20. System for automated processing of a set of images for their publication on a digital medium characterized in that it comprises a first server (SERVi) making it possible to carry out the steps of the method according to any one of claims 1 to 15 and a second server (SERV3) making it possible to carry out the steps of the method according to any one of claims 16 to 19, said processed images being organized in the form of a sequence of ordered images.

Images