EP4285338A1 - System mit einer geschirrspülmaschine, verfahren und computerprogrammprodukt - Google Patents
System mit einer geschirrspülmaschine, verfahren und computerprogrammproduktInfo
- Publication number
- EP4285338A1 EP4285338A1 EP22700619.4A EP22700619A EP4285338A1 EP 4285338 A1 EP4285338 A1 EP 4285338A1 EP 22700619 A EP22700619 A EP 22700619A EP 4285338 A1 EP4285338 A1 EP 4285338A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- image
- 2dimg
- dishwasher
- captured
- metadata
- 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
Classifications
-
- A—HUMAN NECESSITIES
- A47—FURNITURE; DOMESTIC ARTICLES OR APPLIANCES; COFFEE MILLS; SPICE MILLS; SUCTION CLEANERS IN GENERAL
- A47L—DOMESTIC WASHING OR CLEANING; SUCTION CLEANERS IN GENERAL
- A47L15/00—Washing or rinsing machines for crockery or tableware
- A47L15/42—Details
- A47L15/4295—Arrangements for detecting or measuring the condition of the crockery or tableware, e.g. nature or quantity
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/10—Image acquisition
- G06V10/12—Details of acquisition arrangements; Constructional details thereof
- G06V10/14—Optical characteristics of the device performing the acquisition or on the illumination arrangements
- G06V10/141—Control of illumination
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/64—Three-dimensional [3D] objects
-
- A—HUMAN NECESSITIES
- A47—FURNITURE; DOMESTIC ARTICLES OR APPLIANCES; COFFEE MILLS; SPICE MILLS; SUCTION CLEANERS IN GENERAL
- A47L—DOMESTIC WASHING OR CLEANING; SUCTION CLEANERS IN GENERAL
- A47L2401/00—Automatic detection in controlling methods of washing or rinsing machines for crockery or tableware, e.g. information provided by sensors entered into controlling devices
- A47L2401/04—Crockery or tableware details, e.g. material, quantity, condition
Definitions
- the present invention relates to a system with a dishwasher, a method for operating a dishwasher and a computer program product.
- Dishwashers which have a camera or the like for capturing an image of a washware receptacle. For example, a load in the dishwasher, soiling of the items to be washed and the like can be determined on the basis of captured images. Such information can be used to optimize a washing program so that a cleaning result can be improved.
- one object of the present invention is to improve the cleaning of items to be washed with a dishwasher.
- a system with a dishwasher in particular one with a domestic dishwasher, is proposed.
- the dishwasher includes an image capture device for capturing a 2D image of a washware receptacle of the dishwasher and washware arranged therein.
- the system includes an image processing device for determining a 3D image of the item to be cleaned on the basis of the captured 2D image and additionally captured metadata of the captured 2D image.
- the dishwasher also includes a control device for carrying out a washing program as a function of the determined 3D image.
- This system has the advantage that there is a three-dimensional image of the items to be washed, with only one image recording device being required for this.
- the three-dimensional image makes it possible to determine various information that is helpful for carrying out the washing program. For example, based on the 3D image, it can be determined whether items to be washed protrude from the items to be washed and can block a spray device, such as a rotatably mounted spray arm, and appropriate measures can be taken. Furthermore, an intensive rinsing zone can be activated or deactivated automatically. Furthermore, when loading the dishwasher, a user can be supported with items to be washed in order to achieve an optimal load. Furthermore, a drying step can be optimized.
- the image acquisition device preferably includes an image sensor with an extended spectral range, which has a sensitivity in the range between 2500 nm-250 nm, for example.
- the image capturing device preferably also includes a lens which is set up for imaging the items to be washed (or the images to be washed if the dishwasher has a plurality of items to be washed) onto the image sensor.
- the image capture device preferably captures the image of the washware receptacle when it is arranged in a predetermined position.
- the image capturing device can also include an illumination unit in order to illuminate the washware receptacle while the image is being captured.
- the dishwasher has only exactly one image acquisition device with an optical beam path and a sensor. In particular, it is therefore not a stereo camera. The complexity of the dishwasher can thus be kept low.
- the image capturing device can be embodied as a stereo camera that captures two 2D images from different viewing angles and on the basis of which the 3D image can be determined.
- the dishwasher comprises a washing compartment with a washing compartment flange running around a loading opening of the washing compartment, the image capturing device being arranged in or on an upper flange section of the washing compartment flange assigned to a ceiling of the washing compartment, and the image capturing device pointing away from the washing compartment and at an angle in the direction of one of Bottom of the rinsing tank has spanned level.
- the rinsing container flange is preferably located outside of a rinsing chamber which is sealed off by means of a sealing device.
- a cavity is provided in or on the washing container flange, in which the image acquisition device is accommodated at least in sections, the cavity being delimited by a transparent element, preferably in the direction of the plane.
- This embodiment ensures that the image capturing device is arranged on the dishwasher in such a way that it does not protrude or is perceived as being annoying.
- a dishwasher that has the washing tub flange with the cavity but is not equipped with an image acquisition device can also be retrofitted comparatively easily.
- the transparent element is designed, for example, as a plastic pane, a mineral glass pane, a glass pane or the like.
- the transparent element preferably ensures sealing of the cavity downwards, in the direction of the plane, so that when the door is opened after a washing program, steam from the washing chamber cannot enter the cavity.
- a corresponding sealing device is provided for this purpose, for example.
- the transparent element is preferably transparent for the entire spectral range that is captured by the image capturing device. However, in embodiments it can also be provided that the transparent element is transparent only selectively for certain areas of the spectral range. Then the transparent element additionally acts as a filter.
- the imaging device is arranged entirely within the cavity, with the cavity being sealed as a whole. This means that in particular no moisture can penetrate into the cavity, which is advantageous for the service life of the image acquisition device.
- the image capturing device preferably has a vertical angle of view of at least 90° and a horizontal angle of view of at least 120°. This ensures that the item to be washed can be recorded as a whole without individual areas, for example edge areas, of the item to be washed not being visible on the recorded image.
- the respective angle of view is preferably provided by corresponding optical elements. One can also speak of an ultra-wide-angle lens.
- the image processing device can be implemented in terms of hardware and/or software.
- the image processing device can be embodied, for example, as a computer or as a microprocessor.
- the image processing device can be embodied as a computer program product, as a function, as a routine, as part of a program code or as an executable object.
- the image processing device preferably includes a neural network which is set up to generate the 3D image.
- the image processing device can have a number of neural networks, which can each carry out a partial step in the determination of the 3D image.
- the image processing device preferably includes a processor and a memory device and is set up to carry out image transformations of the captured 2D image using algorithms, functions and the like.
- the structure of a neural network is as follows.
- the neural network can be divided into three layers or sections, an input layer, one or more hidden layers, and an output layer.
- Each of the layers includes a plurality of individual neurons, it also being possible for different layers to have a different number of neurons.
- the amount of neurons in the input layer depends, for example, on the amount and the format of the data to be processed by the neural network, which can also be referred to as the dimensionality of the data.
- the amount of neurons in the output layer depends on the dimensionality of the output vector.
- the input layer forms the input of the neural network, which receives the data to be processed in the form of an input vector.
- the neurons of the input layer receive the information contained in the input vector and se weights on to the neurons of the first hidden layer.
- the neurons of the first hidden layer receive the weighted signals from the input layer and pass them on to the following hidden layer in a weighted manner.
- the signals are passed through the neural network from layer to layer in this way until the output layer is reached.
- the output vector is tapped off at the neurons of the output layer.
- the amount of hidden layers is basically unlimited, however the computational power needed to operate the neural network scales with the amount of hidden layers.
- the metadata that is captured for a respective captured 2D image includes, for example, various information about a current operating state of the dishwasher, such as an extension state of a washware holder, an opening angle of a door of the dishwasher and the like, as well as operating parameters or settings of the image capture device for Time of image acquisition, such as an exposure time, a sensitivity, an angle of view and the like.
- the metadata thus includes information that is helpful when determining the 3D image, in particular to avoid misinterpretations.
- a neural network must be trained before it can be deployed. There are different training methods for this.
- supervised learning an input vector is supplied to the neural network, with the corresponding optimum output vector also being predetermined. Based on the difference between the output vector generated by the neural network and the optimal output vector, the neural network is optimized. For this purpose, the weights with which the individual neurons are linked are adjusted.
- unsupervised learning the optimal starting vector is not specified, but instead an optimization function is evaluated, for example, and the neural network is optimized accordingly.
- GAN generative adversarial network
- the determined 3D image can also be referred to as a 3D model of the item to be washed.
- the 3D image includes depth information that is not included in the 2D image.
- the execution of the washing program can be optimized on the basis of the depth information. Optimizing the execution of the washing program includes, in particular, outputting a message to a user of the dishwasher in order to signal a suboptimal arrangement of items to be washed, so that the user can optimize the arrangement, for example. This can preferably be done under the guidance of the dishwasher.
- parameters that relate to the loading of the items to be washed with washing liquor by means of one or more hydraulic systems can be optimized. For example, a higher pressure can be used for deep pots or glasses in order to also clean these dishes in depth.
- the image processing device comprises a first neural network that is set up to generate a latent vector depending on the captured 2D image and the metadata of the captured 2D image, and a second neural network that is set up to do so , to determine a 3D image of the washware receptacle as a function of the generated latent vector.
- the first neural network determines a latent vector based on the captured 2D image and the captured metadata.
- the latent vector thus forms the output vector of the first neural network.
- the latent vector can be understood as an encoding of the information contained in the captured 2D image and metadata.
- the latent vector comprises a significantly reduced amount of data compared to the captured 2D image and the metadata.
- the first neural network is designed in particular as a CNN (convolutional neural network).
- the second neural network uses the latent vector to determine a 3D image of the item to be washed.
- the second neural network is designed in particular as a GAN that has been trained in a corresponding training method.
- Corresponding training data must be available for the training of a respective neural network. For example, a large number of 2D images with corresponding metadata and the latent vectors to be generated from them must be specified. Furthermore, the 3D images belonging to the latent vectors must be known.
- the size of the training data set preferably comprises at least one hundred, preferably several hundred, preferably at least one thousand, preferably several thousand sets, each set comprising a 2D image with metadata, the corresponding latent vector and the corresponding 3D image.
- the latent vector comprises a representation of the image data contained in the captured 2D image and the metadata, with a data quantity of the latent vector at most 1/100, preferably at most 1/500, preferably at most 1/1000, further preferred at most 1/10,000 of a data set of the captured 2D image.
- This embodiment enables the 3D image to be generated quickly on the basis of the latent vector, since comparatively little data has to be processed by the second neural network in order to generate the 3D image.
- the metadata of the captured 2D image includes information regarding the washware receptacle visible on the captured image and/or a section of the washware receptacle visible on the captured 2D image.
- the dishwasher comprises a plurality of items to be washed, such as a lower and an upper item to be washed.
- the information can be determined, for example, on the basis of an optical identification of the item to be washed, which can be seen on the 2D image, using a corresponding algorithm.
- the dishwasher can have a separate sensor that is set up to capture this information and to output it as metadata for the captured image.
- the metadata of the captured 2D image includes information regarding a height adjustment of the washware receptacle at the time the 2D image was captured.
- the dishwashing machine can include a washware receptacle whose height is adjustable.
- the items to be washed can be adapted in particular to accommodate items to be washed of different sizes. Since the distance from the image acquisition device to the washware holder is helpful information for determining the 3D image, the current height setting of the washware holder is preferably recorded as metadata for the 2D image.
- the dishwasher has a corresponding sensor that is set up to detect the height setting of the items to be washed.
- the sensor can be designed as an image processing unit.
- the metadata of the captured 2D image includes information relating to an angle between the image capturing device and the washware holder.
- the dishwasher includes an illumination device for illuminating the washware receptacle and the metadata of the 2D image includes information regarding an operating mode of the illumination device at the time the 2D image was captured.
- the lighting device supports the acquisition of the 2D image, so that in particular a signal-to-noise ratio of the image is better than a predetermined minimum value.
- the lighting device can emit a narrow-band spectrum or emit a broad-band spectrum.
- the lighting device can have an intensity in a spectral range outside the visible spectral range.
- the lighting device can be arranged with a parallax relative to the image capturing device in relation to the wash item holder, so that the image capturing device records illuminated areas of the wash item holder and areas that are shaded by items to be washed.
- Information on the three-dimensional shape of the items to be washed and the receptacle for the items to be washed can be derived on the basis of the shadow cast.
- the metadata include, in particular, information relating to an illumination intensity, an illumination spectrum, and a parallax between the illumination device and the image acquisition device.
- the dishwasher comprises a plurality of spatially separated lighting devices for illuminating the washware receptacle and the metadata of the 2D image includes information regarding a respective operating mode of the plurality of lighting devices at the time the 2D image was captured.
- different lighting states can advantageously be implemented, with a respective lighting state providing additional information for determining the 3D image.
- the determination of the 3D image can thus advantageously take place with increased accuracy and reliability.
- the image capture device is set up to capture multiple 2D images of the washware receptacle in different operating modes of the multiple lighting devices, with the respective metadata of the multiple 2D images providing information regarding the respective operating mode of the multiple lighting devices at the time the respective 2D - Include image, and wherein the image processing device is set up to determine a 3D image depending on the plurality of 2D images and the respective metadata.
- the first neural network is set up to generate exactly one latent vector depending on the plurality of captured 2D images and the respective metadata, or the first neural network is set up to generate a latent vector for each of the plurality of captured 2D images generate, the second neural network being set up to determine a 3D image of the washware receptacle as a function of the plurality of latent vectors generated.
- control device is set up to block a movably mounted spray device of the dishwasher by in the depending on the determined 3D image of the washware receptacle To determine dishwashing arranged dishes and to carry out the washing program depending on the identified blockage.
- a message is preferably output to a user of the dishwasher, so that the user can remove the blockage by removing the items to be washed that are blocking the spray device.
- control device is set up to determine a mass of the washware arranged in the washware holder as a function of the determined 3D image of the washware holder and to carry out the washing program depending on the determined mass.
- the mass of items to be washed has an influence in particular on the drying of the items to be washed and can therefore be taken into account when determining the optimum parameters for the drying step.
- a material of the items to be washed can be taken into account here, which can also be derived from the 2D image, for example.
- control device is set up to determine a liquid collection point of the washware arranged in the washware holder depending on the determined 3D image of the washware holder and to carry out the washing program depending on the determined liquid collection point.
- the liquid collection point is in particular a depression in items to be washed, in which liquid collects and cannot drain away. For example, an indication is issued to the user so that he arranges the items to be washed differently so that the sink is not present. Furthermore, drying parameters can be adjusted so that the liquid in the sink can also dry out.
- the dishwasher has a hydraulic system that includes at least one intensive wash zone, the control device being set up to selectively activate the at least one intensive wash zone depending on the determined 3D image of the washware receptacle.
- the intensive rinsing zone is fluidically coupled to the hydraulic system by means of a switchable valve, for example, with the control device opening the switchable valve to activate the intensive rinsing zone.
- the intensive-rinse zone can also be formed by means of a spray device that can be positioned and/or aligned by a motor, with the control device forming the intensive-rinse zone on the basis of the 3D image in such a way that items to be washed that require intensive treatment, such as pots or pans, are in the intensive-rinse zone.
- the system comprises a device which is external to the dishwasher and which integrates the image processing device, in particular the first neural network and/or the second neural network, the dishwasher having a communication unit which is used to transmit the captured 2D image and the metadata to the external device and is set up to receive the 3D image of the item to be cleaned from the external device.
- the external device is preferably in the form of a server or the like, which has high computing power. Therefore, the 3D image can be determined with complex algorithms and/or using complex neural networks.
- the communication unit can be coupled to the external device by means of a network, for example.
- the network includes in particular a cellular network, a WLAN, the Internet and/or another wireless or wired data network.
- the external device includes, for example, only the first neural network and/or only the second neural network, with the respective other neural network being integrated in particular in the dishwasher.
- the image processing device is thus arranged in a distributed manner.
- the communication unit is accordingly set up to receive the latent vector and/or to transmit the latent vector.
- the method has the same advantages as explained for the system according to the first aspect.
- the method is preferably carried out with the dishwasher of the system according to the first aspect.
- a computer program product which comprises instructions which, when the program is executed by a computer, cause the latter to execute the method described above.
- a computer program product such as a computer program means
- a server in a network, for example, as a storage medium such as a memory card, USB stick, CD-ROM, DVD, or in the form of a downloadable file. This can be done, for example, in a wireless communication network by transferring a corresponding file with the computer program product or the computer program means.
- FIG. 1 shows a schematic perspective view of an embodiment of a domestic dishwasher
- FIG. 2 schematically shows an embodiment for determining a 3D image based on a 2D image and metadata
- FIG. 3 shows a schematic view of a further embodiment of a domestic dishwasher
- Fig. 4 shows a schematic flowchart as an example for determining a latent vector
- FIG. 5 shows a schematic block diagram of an arrangement for training a neural network
- FIG. 6 shows a schematic block diagram of a system with a dishwasher and an external unit
- FIG. 7 shows a schematic block diagram of an exemplary embodiment of a method for operating a dishwasher.
- the domestic dishwasher 1 comprises a washing compartment 2, which can be closed by a door 3, which is in particular watertight.
- a sealing device can be provided between the door 3 and the washing container 2 .
- the washing container 2 is preferably cuboid.
- the rinsing container 2 can be housing the household dishwasher 1 can be arranged.
- the washing compartment 2 and the door 3 can form a washing chamber 4 for washing items to be washed.
- the door 3 is shown in FIG. 1 in its open position.
- the door 3 can be closed or opened by pivoting about a pivot axis 5 provided at a lower end of the door 3 .
- a loading opening 6 of the washing compartment 2 can be closed or opened.
- the washing compartment 2 has a base 7, a cover 8 arranged opposite the base 7, a rear wall 9 arranged opposite the closed door 3 and two side walls 10, 11 arranged opposite one another.
- the bottom 7, the top 8, the rear wall 9 and the side walls 10, 11 can be made of a stainless steel sheet, for example.
- the bottom 7 can be made of a plastic material.
- the household dishwasher 1 also has at least one washware holder 12 to 14 .
- washware holders 12 to 14 can be provided, with washware holder 12 being a lower washware holder or a lower rack, washware holder 13 being an upper washware holder or an upper rack and washware holder 14 being a cutlery drawer.
- the wash items receptacles 12 to 14 are arranged one above the other in the wash tub 2 .
- Each washware holder 12 to 14 can be moved either into or out of the washing compartment 2 .
- each washware receptacle 12 to 14 can be pushed or moved into the washing compartment 2 in an insertion direction E and pulled out or moved out of the washing compartment 2 in an extension direction A counter to the insertion direction E.
- An image capturing device 110 is arranged in an upper area of a washing container flange running around the loading opening 6 of the washing container 2 .
- the image acquisition device 110 is in the form of a digital camera.
- the digital camera 110 is placed, in particular, centrally in relation to the loading opening 6 and has, for example, a wide-angle lens (not shown) that enables each of the washware receptacles 12 - 14 to be viewed in a respective 2 D -Image 2DIMG (see Fig. 2, 4 or 6) to capture completely.
- An image processing device 120 and a control device 140 are also arranged on the door 3 . These are shown separately from one another here, but they can also be integrated together in one element. In addition, the arrangement on the door 3 is only an example.
- the image processing device 120 is used to determine a 3D image 3DIMG of a respective item of washware 12 - 14 (see Fig. 2, 5 or 6) on the basis of the recorded 2D image 2DIMG and additionally recorded metadata META (see Fig. 2 or 6) of the captured 2D image 2DIMG set up. This is explained in detail below with reference to FIGS. 2-6.
- the control device 140 is set up to carry out a flushing program as a function of the determined 3D image 3DIMG. .
- FIG. 2 schematically shows an exemplary embodiment for determining a 3D image 3DIMG based on a 2D image 2DIMG and metadata META, this example having two separate steps.
- the 2D image 2DIMG shows a washware receptacle 12 with washware 21 - 25 arranged therein.
- the 2D image 2DIMG was captured, for example, by the image acquisition device 110 (see FIG. 1) when the washware receptacle 12 was fully shifted out of the washing compartment 4 .
- the meta data META include, for example, the information that the washware receptacle 12 was completely relocated out of the washing compartment 4 at the time the image was captured.
- the metadata META preferably also includes the information which can be seen from a plurality of washware images 12 - 14 on the 2D image 2DIMG, a height setting of the visible washware image 12 and/or an angle of the image acquisition device 110 and the washware image 12.
- the metadata META a timestamp and settings of the image capture device
- a sensitivity e.g. an ISO value or gain value
- a focal length e.g. an angle of view
- an additional lighting device e.g. an additional lighting device
- the metadata META includes in particular information relating to an operating state of the lighting devices 111, 112, such as an illumination intensity, an illumination spectrum and the like.
- the 2D image 2DIMG and the metadata META are combined into a latent vector LVEC.
- the latent vector LVEC includes the information necessary or relevant for determining the 3D image 3DIMG.
- the latent vector LVEC has an amount of data that is at most 1/100, preferably preferably at most 1/500, preferably at most 1/1000, more preferably at most 1/10,000, of the data volume of the captured 2D image 2DIMG.
- the generation of the latent vector LVEC by a neural network 125 is explained in detail with reference to FIG.
- the 3D image 3DIMG is determined on the basis of the latent vector LVEC. This is preferably done by means of a second neural network 130 (see FIG. 5 or 6), which is designed in particular as a GAN.
- Fig. 3 shows a schematic view of a further embodiment of a household dishwasher 1.
- Fig. 3 shows in particular a front view of the household dishwasher 1 with the door 3 open.
- the household dishwasher 1 of Fig. 3 can, like the one based on Fig. 1 explained domestic dishwasher 1 and have all the features explained there, even if they are not shown in Fig. 3.
- the household dishwasher 1 has two lighting devices 111, 112 in addition to the centrally arranged image capturing device 110, which are arranged next to the image capturing device 110.
- the lighting devices 111, 112 include, for example, a flash unit and/or an LED lighting unit.
- the lighting devices 111, 112 can be designed identically or differently from one another. In embodiments, only a single lighting device and/or more than two lighting devices can be provided.
- the lighting devices 111, 112 are set up in particular to illuminate the washware receptacles 12, 13 when the image capturing device 110 captures a 2D image 2DIMG (see FIG. 2, 4 or 6).
- the offset arrangement of image capturing device 110 and lighting devices 111, 112 results in particular in a parallax between the image capturing device 110 and the lighting devices 111, 112.
- the captured 2D image 2DIMG can therefore also include shaded areas, which makes it possible to determine the 3D image 3DIMG ( see Fig. 2, 5 or 6).
- the two lighting devices 111, 112 are preferably set up to illuminate the respective washware receptacle 12, 13 with different spectra, for example with different colors.
- the image capturing device captures 110 in particular a 2D image 2DIMG, which includes an increased information density with regard to the three-dimensional shape of the items to be washed 21 - 25 arranged in the items to be washed 12, 13 (see FIG. 2), since shaded areas can result from two directions of illumination.
- the metadata META (see FIG. 2) of the captured 2D image 2DIMG includes, in particular, information on a relative arrangement of the image capture device 110 and the lighting devices 111, 112 and on the operating state of each of the lighting devices 111, 112 at the time of image capture.
- the image capturing device 110 captures a plurality of 2D images 2DIMG for different operating modes of the lighting devices 111, 112, in particular for different illumination spectra, in order to capture additional image information relating to the items to be washed 21 - 25 arranged in the items to be washed 12, 13.
- FIG. 4 shows a schematic flowchart as an example for determining a latent vector LVEC based on a 2D image 2DIMG and associated metadata META.
- the latent vector LVEC is generated by a trained neural network 125 which comprises a number of functional layers.
- these are a first filter layer CL1 (convolutional layer), a first aggregation layer PL1 (pooling layer), a second filter layer CL2, a second aggregation layer PL2 and a multi-layer perceptron MLP (multi-layered perceptron).
- the two filter layers CL1, CL2 are set up to determine a number of so-called feature maps FM on the basis of the respective input data.
- a respective filter layer CL1, CL2 includes, in particular, a plurality of filters, with each filter generating a feature map FM.
- the generated feature maps FM correspond in particular to a more compact and abstract representation of the input information. The amount of data is therefore reduced by using the filter layers CL1, CL2.
- the aggregation layers PL1, PL2 further reduce the amount of data by a respective aggregation layer PL1, PL2, for example, only the most relevant information of the input data (for the first aggregation layer PL1, for example, the Feature maps FM of the first filter layer CL1) passes on (for example in the form of a so-called "max-pooling layer").
- the number and/or the order of the filter layers CL1 , CL2 and the aggregation layers PL1 , PL2 can deviate from that shown in FIG. 4 in further embodiments.
- the multi-layer perceptron MLP is designed in particular as a fully connected neural network (fully connected layer).
- the multi-layer perceptron MLP generates the latent vector LVEC based on the output data of the second aggregation layer PL2.
- Fig. 5 shows a schematic block diagram of an arrangement for training a neural network 130.
- the neural network 130 is trained, for example, on the basis of a latent vector LVEC, which, for example, as described with reference to FIG 2DIMG (see Fig. 2, 4 or 6) and associated metadata META (see Fig. 2, 4 or 6) was generated, a 3D image 3DIMG is generated.
- the neural network 130 can in particular be part of the image processing device 120 .
- FIG. 5 corresponds in particular to a GAN (generative adversarial network).
- a GAN is an unsupervised learning method in which two neural networks 130, 132 compete against each other to a certain extent.
- Latent vectors LVEC on the one hand and 3D images 3DIMG* on the other hand are specified as training data.
- the specified 3D image 3DIMG* here shows the reality and the corresponding latent vector LVEC is the latent vector LVEC derived on the basis of the corresponding 2D image 2DIMG.
- the neural network 130 is also referred to as a generator since it generates an artificial 3D image 3DIMG.
- the generated 3D image 3DIMG and the real 3D image 3DIMG* are supplied to the neural network 132 .
- the neural network 130 is also referred to as a discriminator.
- the discriminator 130 decides which of the two supplied images 3DIMG*, 3DIMG is the real image and outputs a result R accordingly.
- the generator 130 and the discriminator 132 are in contention: the generator 130 attempts to generate a 3D image 3DIMG which the discriminator 132 mistakes for the real 3D image 3DIMG*.
- the generator 130 is enabled to generate a 3D image 3DIMG that corresponds to an actual three-dimensional view of the washware receptacle 12, 13, 14 (see FIG. 1 or 3).
- the trained generator 130 is part of the image processing device 120, for example.
- the dishwasher 1 is designed in particular as a domestic dishwasher and can have the features of the domestic dishwashers 1 explained in FIGS.
- the domestic dishwasher 1 has a communication unit 160 in this example.
- the communication unit 160 is set up to establish a communication connection with an external device 200 .
- the communication connection can be established wirelessly, for example by means of WLAN, Bluetooth and/or a mobile radio standard according to 2G, 3G, 4G and/or 5G.
- the communication connection can also be established by wire.
- the external device 200 is designed in particular as a computer or server that can be reached via the Internet.
- the external device 200 also has a communication unit 202 .
- the external device 200 integrates the image processing device 120.
- the dishwasher 1 transmits the captured 2D image 2DIMG of the washware receptacle 12, 13, 14 (see Fig. 1 or 3) and the associated metadata META by means of the communication unit 160 to the external device 200.
- the external device 200 includes the image processing device 120, which a first neural network 125, in particular a CNN (convolutional neural network), and a second neural network 130, in particular a GAN.
- the image processing device determines the 3D image 3DIMG of the washware receptacle 12, 13, 14 and transmits the determined 3D image 3DIMG to the domestic dishwasher 1.
- the dishwasher 1 in FIG. 6 also has the image processing device 120 . Then, for example, the processing could take place locally, ie using the image processing device 120 of the dishwasher 1 , or externally, ie using the image processing device 120 of the external device 200 . This can be advantageous if, for example, there is a connection problem with the communication link.
- the dishwasher 1 includes, for example, the first neural network 125 and the external device 200 includes the second neural network 130.
- the latent vector LVEC is transmitted from the dishwasher 1 to the external device 200.
- the dishwasher 1 comprises, for example, the second neural network 130 and the external device 200 comprises the first neural network 125.
- the latent vector LVEC from the external device 200 is received by the dishwasher 1 here.
- FIG. 7 shows a schematic block diagram of an exemplary embodiment of a method for operating a dishwasher 1, for example the domestic dishwasher of FIG. 1 or FIG. 3.
- a 2D image 2DIMG (see Fig. 2, 4 or 6) of a washware receptacle 12, 13, 14 (see Fig. 1, 2 or 3) of the dishwasher 1 and washware 21-25 arranged therein (see Fig. 2) and metadata META (see Fig. 2 or 3) to the 2D image 2DIMG.
- a 3D image 3DIMG (see FIG. 2, 5 or 6) of the washware receptacle 12, 13, 14 is determined on the basis of the recorded 2D image 2DIMG and the recorded material data META. This takes place, for example, as described above with reference to FIGS.
- a flushing program is carried out as a function of the determined 3D image 3DIMG. For example, based on the 3D image 3DIMG, it can be determined whether items to be washed 21-25 protrude from the items to be washed 12, 13, 14 and can block a spray device, such as a rotatably mounted spray arm, and appropriate measures can be taken. Furthermore, an intensive rinsing zone can be activated or deactivated automatically. Furthermore, a user can be supported when loading the dishwasher 1 with items to be washed 21 - 25 in order to achieve optimal loading. Furthermore, a drying step can be optimized.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| DE102021200715.5A DE102021200715A1 (de) | 2021-01-27 | 2021-01-27 | System mit einer Geschirrspülmaschine, Verfahren und Computerprogrammprodukt |
| PCT/EP2022/050623 WO2022161778A1 (de) | 2021-01-27 | 2022-01-13 | System mit einer geschirrspülmaschine, verfahren und computerprogrammprodukt |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| EP4285338A1 true EP4285338A1 (de) | 2023-12-06 |
Family
ID=80001331
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| EP22700619.4A Pending EP4285338A1 (de) | 2021-01-27 | 2022-01-13 | System mit einer geschirrspülmaschine, verfahren und computerprogrammprodukt |
Country Status (5)
| Country | Link |
|---|---|
| US (1) | US12495948B2 (de) |
| EP (1) | EP4285338A1 (de) |
| CN (1) | CN116762104A (de) |
| DE (1) | DE102021200715A1 (de) |
| WO (1) | WO2022161778A1 (de) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP4509031A1 (de) | 2023-08-16 | 2025-02-19 | Miele & Cie. KG | Verfahren und vorrichtung zum erkennen eines innenraums eines reinigungsgeräts und reinigungsgerät |
| DE102023211364A1 (de) * | 2023-11-15 | 2025-05-15 | BSH Hausgeräte GmbH | Geschirrspüler und Verfahren zum Erkennen eines Objekts darin |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102004035847A1 (de) | 2004-07-23 | 2006-03-23 | BSH Bosch und Siemens Hausgeräte GmbH | Verfahren zur Erkennung der Spülgutbeladung und Geschirrspülmaschine |
| US10809685B2 (en) | 2019-03-18 | 2020-10-20 | Midea Group Co., Ltd. | Dishwasher with cloud connected cameras |
| DE102019111848B4 (de) | 2019-05-07 | 2023-12-21 | Illinois Tool Works Inc. | System zur optischen Spülguterkennung bei Spülmaschinen, Verfahren zur optischen Spülguterkennung, Spülmaschine mit einem optischen Spülguterkennungssystem sowie Verfahren zum Betreiben einer solchen Spülmaschine |
| KR102745358B1 (ko) | 2019-08-06 | 2024-12-20 | 엘지전자 주식회사 | 인공지능 장치를 이용한 식기 세척 방법 및 이를 위한 장치 |
| US11439292B2 (en) * | 2019-11-04 | 2022-09-13 | Midea Group Co. Ltd. | System and method for recommending object placement |
-
2021
- 2021-01-27 DE DE102021200715.5A patent/DE102021200715A1/de active Pending
-
2022
- 2022-01-13 CN CN202280011911.7A patent/CN116762104A/zh active Pending
- 2022-01-13 EP EP22700619.4A patent/EP4285338A1/de active Pending
- 2022-01-13 US US18/265,704 patent/US12495948B2/en active Active
- 2022-01-13 WO PCT/EP2022/050623 patent/WO2022161778A1/de not_active Ceased
Also Published As
| Publication number | Publication date |
|---|---|
| US20240041296A1 (en) | 2024-02-08 |
| CN116762104A (zh) | 2023-09-15 |
| DE102021200715A1 (de) | 2022-07-28 |
| WO2022161778A1 (de) | 2022-08-04 |
| US12495948B2 (en) | 2025-12-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| DE102015102694B4 (de) | Verfahren zur Ermittlung der Drehzahl eines Sprüharms einer Geschirrspülmaschine | |
| WO2015117905A1 (de) | 3d-bildanalysator zur blickrichtungsbestimmung | |
| EP4285338A1 (de) | System mit einer geschirrspülmaschine, verfahren und computerprogrammprodukt | |
| EP3654818B1 (de) | Haushaltsgeschirrspülmaschine und verfahren zum betreiben einer haushaltsgeschirrspülmaschine | |
| WO2019015996A1 (de) | Haushaltsgeschirrspülmaschine und verfahren zum betreiben einer haushaltsgeschirrspülmaschine | |
| DE102013110642A1 (de) | Gargerät mit Kamera und Verfahren zur Einschuberkennung | |
| WO2006015934A1 (de) | Verfahren zur erkennung der spülgutbeladung und geschirrspülmaschine | |
| EP3654815A1 (de) | Haushaltsgeschirrspülmaschine und verfahren zum betreiben einer haushaltsgeschirrspülmaschine | |
| DE102020112205A1 (de) | Geschirrspülmaschine mit einer Bilderfassungseinrichtung und Verfahren zum Betreiben einer derartigen Geschirrspülmaschine | |
| DE102018202092A1 (de) | Verfahren zum Betreiben einer Haushaltsgeschirrspülmaschine, Haushaltsgeschirrspülmaschine und System | |
| EP4106596B1 (de) | System mit einer geschirrspülmaschine und verfahren zum betreiben einer geschirrspülmaschine | |
| EP3654817B1 (de) | Haushaltsgeschirrspülmaschine und verfahren zum betreiben einer haushaltsgeschirrspülmaschine | |
| DE102019110795A1 (de) | Geschirrspülmaschine mit einer Bilderfassungseinrichtung und Verfahren zum Betreiben einer derartigen Geschirrspülmaschine | |
| WO2022002689A1 (de) | System mit einer geschirrspülmaschine und verfahren zum betreiben einer geschirrspülmaschine | |
| EP3530172B1 (de) | Spülmaschine mit beladungserkennung | |
| DE102018209973A1 (de) | Geschirrspülmaschine, Verfahren zum Betreiben einer Geschirrspülmaschine und Computerprogrammprodukt | |
| EP4167824A1 (de) | System mit einer geschirrspülmaschine, verfahren zum betreiben einer geschirrspülmaschine und computerprogrammprodukt | |
| EP3932282A1 (de) | Spülgerät mit bilderfassungseinrichtung | |
| DE102020211543A1 (de) | System mit einer Geschirrspülmaschine, Verfahren zum Betreiben einer Geschirrspülmaschine und Computerprogrammprodukt | |
| EP4509031A1 (de) | Verfahren und vorrichtung zum erkennen eines innenraums eines reinigungsgeräts und reinigungsgerät | |
| BE1031766B1 (de) | Verfahren und Steuervorrichtung zum Anpassen eines Reinigungsvorgangs in einem Reinigungsgerät, Reinigungsgerät | |
| BE1029768B1 (de) | Verfahren zum Überwachen eines Beladungszustands eines Spülgeräts, Vorrichtung und Spülgerät | |
| CN115456863B (zh) | 一种图像生成方法及装置、图像中干扰的去除方法、炊具 | |
| BE1030060B1 (de) | Bedienvorrichtung und Verfahren zum Bedienen eines Reinigungsgeräts und Reinigungsgerät | |
| EP4708224A1 (de) | Verfahren zum betreiben eines haushaltsgeräts, steuereinheit und haushaltsgerät |
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: 20230828 |
|
| 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) | ||
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: EXAMINATION IS IN PROGRESS |
|
| 17Q | First examination report despatched |
Effective date: 20250612 |