Method of selecting one or more lighting devices from a database of lighting devices

Abstract

A method (500) of selecting one or more lighting devices (200) from a database (110) of lighting devices (200) based on a target image (150) depicting a first space (152), such as a room (152), the method (500) comprising: inputting (502) the target image (150) into an identification and classification system (120) configured to identify and classify lighting-related attributes in images. Analyzing (504), using the identification and classification system (120), the target image (150) to identify lighting-related attributes (LA) in the target image (150), and to classify one or more lighting-related attributes (LA), among the identified lighting-related attributes (LA), as light interference-related attributes (IA) pertaining to transparent or reflective surfaces depicted in the target image (150). Selecting (506), by a processing device (130), one or more lighting devices (200) from the database (110) based on a subset (LA´) of the identified lighting-related attributes (LA), the subset (LA´) comprising the identified lighting-related attributes (LA) except the light interference- related attributes (IA), wherein the selected one or more lighting devices (200) having a light emission which recreates the lighting-related attributes (LA) of the subset (LA´) when used to illuminate a second space.

Description

[0001] 2024PF80406

[0002] 1

[0003] METHOD OF SELECTING ONE OR MORE LIGHTING DEVICES FROM A DATABASE OF LIGHTING DEVICES

[0004] FIELD OF THE INVENTION

[0005] The present invention generally relates to a method of selecting one or more lighting devices from a database of lighting devices based on a target image depicting a space, such as a room.

[0006] BACKGROUND OF THE INVENTION

[0007] Lighting products used to illuminate a room are paramount when it comes to creating desired illumination conditions and hence a desired atmosphere in the room. If a room is illuminated using some lighting products, the illumination conditions and atmosphere created in the room can change drastically if other lighting products are used to illuminate the very same room. Thus, it is important to use carefully selected lighting products in a room if one wants to create certain illumination conditions and atmosphere in the room, although other characteristics such as colors, style and furniture also plays an important role in atmosphere creation. A photo of a room is a great way of illustrating the illumination conditions in a room, i.e. how the room is illuminated using lighting products. Further, a photo of a room is a great way of illustrating the atmosphere of the room.

[0008] Given this, it has become increasingly popular to use photos as input in various product recommendation systems, such as light product recommendation systems or light product selection systems. In such light product recommendation systems relevant lighting products can be recommended or selected from a photo used as an input. Typically, such systems are powered by some form computer implemented method used to analyze an inputted photo. Thus, a user utilizing such a system can from a photo of a room receive a recommendation of lighting products, where the lighting products are selected so as to resemble the illumination conditions and ambiance in the room of the photo. In this way, a uses can take a photo of a room he or she likes, or browse, e.g. the internet, for photos of rooms he or she likes.

[0009] However, when utilizing light product recommendation systems or light product selection systems of the above type, the product recommendations or product2024PF80406

[0010] 2

[0011] selections generated by the systems tend to be unpredictable and hence unreliable. In other words, lighting product recommendation systems tends to give faulty or less relevant product recommendations or selections resulting in that the desired illumination conditions and ambiance cannot be resembled by the recommended or selected lighting products to satisfaction.

[0012] SUMMARY OF THE INVENTION

[0013] It is an object of the present invention to provide method of selecting one or more lighting devices from a database of lighting devices based on a target image depicting a first space, such as a room, which solves or at least alleviates the above drawbacks of the prior art.

[0014] These and other objects may be achieved by method of selecting one or more lighting devices from a database of lighting devices based on a target image depicting a first space, such as a room, in accordance with claim 1. Embodiments of the present invention are defined in the dependent claims.

[0015] According to an aspect of the present invention, there is provided a method of selecting one or more lighting devices from a database of lighting devices based on a target image depicting a first space, such as a room, the method comprising: inputting the target image into an identification and classification system configured to identify and classify lighting-related attributes in images, analyzing, using the identification and classification system, the target image to identify lighting-related attributes in the target image, and to classify one or more lighting-related attributes, among the identified lighting-related attributes, as light interference-related attributes pertaining to transparent or reflective surfaces depicted in the target image, selecting, by a processing device, one or more lighting devices from the database based on a subset of the identified lighting-related attributes, the subset comprising the identified lighting-related attributes except the light interference-related attributes, wherein the selected one or more lighting devices having a light emission which recreates the lighting-related attributes of the subset when used to illuminate a second space.

[0016] The present invention is based on the realization that the selection of one or more lighting devices based on a target image may be significantly improved by disregarding certain lighting-related attributes identified in the target image. More specifically, the selection of one or more lighting devices based on the target image may be significantly improved by disregarding one or more lighting-related attributes among identified lighting-2024PF80406

[0017] 3

[0018] related attributes which are attributable to light-interference in the target image. In this way, adequate lighting devices having a light emission which recreates the lighting-related attributes used for the actual selection when used to illuminate a second space may be selected.

[0019] Hence, the present invention is based on the idea that a target image is inputted into an identification and classification system configured to identify and classify lighting-related attributes in images, where the target image is analyzed by the machine learning identification and classification system to identify lighting-related attributes in the target image. One or more of the identified lighting-related attributes are then classified as light interference-related attributes pertaining to transparent or reflective surfaces depicted in the target image by the identification and classification system. In the following selection, of one or more lighting devices from the database, the lighting-related attributes classified as light interference-related attributes are disregarded. In this way, light related attributes which are attributable to light effects not directly linked to lighting devices present in the space of the target image may be disregarded when selecting the one or more light lighting devices from the database. This further enables a selection of one or more lighting devices having a light emission which recreates the lighting-related attributes attributable to actual lighting devices in the target image when used to illuminate a second space. Put differently, one or more lighting devices having a light emission which recreates the lighting conditions of the actual lighting devices depicted in the target image may be selected.

[0020] By “identification and classification system” is here meant any computer or machine implemented system which is adapted to analyze an inputted image to identify and classify lighting-related attributes in the inputted image. The identification and classification system may thus comprise image processing capabilities. Such image processing capabilities may in practice be optimized to identify and classify lighting-related attributes in an inputted image.

[0021] By “lighting-related attributes” is here meant any elements, characteristics, portions or regions in the target image pertaining or related to lighting devices depicted in the target image. The lighting-related attributes may thus be attributable to lighting devices in the space being directly depicted. The lighting-related attributes may be attributable to lighting devices outside of the space but depicted in the target image. The lighting-related attributes may be attributable to reflections of lighting devices depicted in the target image. The lighting-related attributes may be attributable to glare of lighting devices depicted in the target image.2024PF80406

[0022] 4

[0023] In some embodiments, the classification and identification system may comprise an artificial intelligence, Al, model trained to identify and classify lighting-related attributes in images, and wherein the analyzing may comprise using the Al model. Thus, the Al model may be used to analyze the target image to identify lighting-related attributes in the target image, and to classify one or more lighting-related attributes, among the identified lighting-related attributes, as light interference-related attributes pertaining to transparent or reflective surfaces depicted in the target image.

[0024] By the Al model being “trained to identify and classify lighting-related attributes in images” is here meant that the Al model is capable of identifying identify and classify lighting-related attributes in images. Thus, the Al model must not be trained using images similar to the target image but is capable of identifying lighting-related attributes in images and to classify the same.

[0025] In some embodiments, the Al model may be machine learning model.

[0026] In some embodiments, the Al model may be a neural network model, such as a large language model, LLM. In this way, a large amount of data or background data may be used when identifying and classifying lighting-related attributes in the target image. The use of a neural network further enables that virtually any motif and style of the target image may be analyzed. The use of an LLM also enables that virtually any motif and style of the target image may be analyzed.

[0027] In some embodiments, an identified light-related attribute originating from one or more of; a depicted light source shining through a window, a depicted light source being reflected in a window, a depicted light source being reflected in a reflective surface, a depicted light reflection in a window, a depicted light shining through a window and a depicted light reflection in a reflective surface may be classified as a light interference-related attribute. In this way, lighting-related attributes not attributable to lighting devices directly depicted in the target image may be classified as light interference-related attributes.

[0028] In some embodiments, the analyzing may further comprise, identifying one or more ambiance-related attributes of the first space. In this way, the ambiance or atmosphere of the first space depicted in the target image may be identified or evaluated and hence subsequently taken into account when selecting the one or more lighting devices from the database.

[0029] In some embodiments, an identified ambience-related attribute may be indicative of, a space type of the first space, a type of furniture present in the first space or an atmosphere type in the first space.2024PF80406

[0030] 5

[0031] In some embodiments, the selecting may further comprise, selecting the one or more lighting devices based on the identified one or more ambience-related attributes, such that a target ambience condition of the first space is mimicked in the second space when the selected one or more lighting devices are used to illuminate the second space. In this way, the one or more lighting devices may be selected in such a way that the one or more lighting devices matches the style and, hence the ambience of the target image. According to an example, one or more lighting devices considered as having a contemporary style may be selected when the target image depicts a first space being considered having a contemporary style. According to an example, one or more lighting devices considered as having a country style may be selected when the target image depicts a first space being considered having a country style.

[0032] In some embodiments, the first space may be of a first space type and wherein the second space is of the first space type. In this way, the one or more lighting devices may be selected in such a way that the one or more lighting devices matches the type of space in the target image.

[0033] In some embodiments, the analyzing may further comprise, classifying one or more lighting-related attributes, among the identified lighting-related attributes, as lighting device-related attributes pertaining to a respective lighting device depicted in the target image.

[0034] In some embodiments, the selecting may further comprise, selecting the one or more lighting devices based on the lighting device-related attributes. In this way, the one or more lighting devices may be selected in such a way that the one or more lighting devices matches the lighting devices depicted in the target image. For example, a table lamp and floor lamp may be selected based on classified lighting device-related attributes attributable to a table lamp and a floor lamp respectively.

[0035] In some embodiments, the method may further comprise, generating, for each lighting device-related attribute, a target feature vector representative of features of the lighting device-related attribute by analyzing characteristics of the lighting device-related attribute. Such feature vector representative of features of the lighting device-related attribute may include features related to a type of lighting device, a color of the lighting device, a brightness of the lighting device, a light temperature of the lighting device, a size of the lighting device, an illumination pattern of the lighting device or similar to give a few nonlimiting examples.2024PF80406

[0036] 6

[0037] In some embodiments, the method may further comprise, generating, for each lighting device-related attribute, a target feature vector representative of features of the lighting device-related attribute and the ambiance-related attributes by analyzing characteristics of the lighting device-related attribute and of the ambiance-related attributes.

[0038] In some embodiments, the database may be a vector type database, wherein each lighting device of the database is associated with a predetermined feature vector, and wherein the selecting one or more lighting devices comprises selecting a lighting device associated with a feature vector similar to the target feature vector by performing a similarity search in the database. According to an example, in this way one or more lighting devices having a certain type, or types may be selected when the target image depicts a first space having one or more lighting devices of the certain type, or types. Thus, if the target image depicts a first space having a small, contemporary, blue table lamp with a bright intense light and large a floor lamp with a warm glow, a similar (or a same model of) table lamp and a similar (or a same model of) floor lamp may be selected.

[0039] In some embodiments, the method may further comprise, generating, by the processing device, a user comprehensible recommendation of lighting devices comprising the one or more selected lighting devices. Thus, such recommendation may be generated and consequently outputted in a format which is comprehensible to a human being. Such user comprehensible recommendation of lighting devices may according to an example include one or more of an image, text, lighting device related product data or similar.

[0040] In some embodiments, the Al model may be neural network model, or a large language model, LLM.

[0041] In some embodiments, the Al model may be neural network model.

[0042] In some embodiments, the Al model may be a large language model, LLM. According to a second aspect of the present invention, there is provided a computer program product comprising computer program code means which, when executed on a computing device having a processing system, cause the processing system to perform all of the steps of the method of the first aspect.

[0043] The above-mentioned features and advantages of the first aspect, when applicable, apply to this second aspect as well. In order to avoid undue repetition, reference is made to the above.

[0044] BRIEF DESCRIPTION OF THE DRAWINGS

[0045] This and other aspects of the present invention will now be described in more2024PF80406

[0046] 7

[0047] detail, with reference to the appended drawings showing embodiments of the present invention.

[0048] Fig. 1 shows an exemplifying block diagram of a system which may be used to implement the method of the present invention.

[0049] Fig. 2 shows a flow chart of a method of selecting one or more lighting devices from a database of lighting devices based on a target image depicting a first space, such as a room according to an exemplifying embodiment of the present invention.

[0050] Fig. 3 shows an example of a target image of a space.

[0051] Fig. 4 shows an example of a target image of another space.

[0052] DETAILED DESCRIPTION

[0053] Fig. 1 schematically shows a block diagram of an exemplifying system 100 which may be used to implement a method 500 of the present disclosure. A schematic flow scheme of an exemplifying method 500 according to the present inventive concept is depicted in Fig. 2. In the following, reference is made to Figs. 1 and 2. Thus, the system 100 of Fig. 1 may be used to implement the method 500 of selecting one or more lighting devices from a database of lighting devices based on a target image depicting a first space, such as a room, as schematically illustrated in Figs 3 and 4 to which reference is also made.

[0054] The depicted system 100 of Fig. 1 comprises a database of 110 of lighting devices 200. That is, the depicted database 110 includes data posts or entries 200: l-200:n. In the following, the lighting devices 200 and the data posts 200:l-200:n will be jointly referred to as 200 wherever suitable for the legibility of the present disclosure. Each data post 200: 1-200:n represents a respective lighting device 200 of a product portfolio of lighting devices 200. The depicted database 110 is a vector type database 110. In the database 110, each lighting device 200 of the database is associated with a predetermined feature vector 202: 1-202:n. To this end, each feature vector 202 of the database 110 includes features representing the respective lighting devices 200. Thus, a feature vector 202 may represent various characteristics of its associated lighting device 200, such as the type of the lighting device, the color of the lighting device, the brightness of the lighting device, the light temperature of the lighting device, the size, i.e. measures, of the lighting device, the illumination pattern of the lighting device or similar. It is, however, to be understood that any type of suitable database 110 may be used to advantage. The database 110 may be a local database, a cloudbased database, distributed database, or a combination thereof to give a few non-limiting examples.2024PF80406

[0055] 8

[0056] The depicted system 100 of Fig. 1 further comprises an identification and classification system 120. The identification and classification system 120 is configured to identify and classify lighting-related attributes in images. In the depicted system 100 of Fig.

[0057] 1, the identification and classification system 120 is embodied in the form of an Al model 120 or machine learning model 120. More specifically, the depicted identification and classification system 120 of Fig. 1 is embodied the form of a neural network model 120. In the following, the wordings Al model 120, machine learning model 120 and neural network model 120 may be used interchangeably. It is however to be noted that the wordings Al model 120 and machine learning model 120 encompasses other machine implemented models 120 apart from neural network models 120. The neural network model 120 is trained to identify and classify lighting-related attributes LA in images. The neural network model 120 is configured to receive a target image 150 as an input, where the target image 150 depicts a first space 152. Typically, the first space 152 may be a room. To this end, the method 500 of selecting one or more lighting devices 200 from the database 110 of lighting devices 200 based on a target image 150 depicting a first space 152, such as a room, as schematically illustrated in Fig. 2, starts with inputting 502 the target image 150 into the neural network model 120 trained to identify and classify lighting-related attributes in images, i.e., the LLM of the system 100 of Fig. 1 according to the present exemplifying disclosure.

[0058] When the identification and classification system 120 is embodied in the form of an Al model 120 or neural network model 120, the Al model 120 or neural network model 120 may be of any suitable type as long as the Al model 120 or neural network model 120 is capable of identifying identify and classify lighting-related attributes LA in images. Such neural network model 120 may for instance include a feedforward neural networks, FNN, a convolutional neural network, CNN, a recurrent neural network, RNN or a generative adversarial networks, GAN, to give a few non-limiting examples. In more general words, such Al model 120 may include machine learning models, deep learning models or generative Al models to give a few more non-limiting examples. However, in the following the depicted identification and classification system 120 of Fig. 1 is exemplified as being a large language model 120, LLM. In the following, the present inventive concept will be described in a context where the identification and classification system 120 is a large language model 120. The identification and classification system 120 may be cloud based or implemented in the cloud. Alternatively, or additionally, the identification and classification system 120 may be implemented using dedicated hardware or software. To this end, the2024PF80406

[0059] 9

[0060] identification and classification system 120 may include one or more of a CPU ("Central Processing Unit"), a DSP ("Digital Signal Processor"), a microprocessor, a microcontroller, an ASIC ("Application-Specific Integrated Circuit"), a combination of discrete analog and / or digital components, or some other programmable logical device, such as an FPGA ("Field Programmable Gate Array").

[0061] The identification and classification system 120, i.e. the LLM of Fig. 1 is further configured to analyze an inputted target image 150 or input target image 150. More specifically, the LLM of Fig. 1 is configured to analyze an inputted target image 150 to identify lighting-related attributes LA in the target image 150. The LLM 120 of Fig. 1 is further configured to classify identified lighting-related attributes LA as light interference-related attributes IA pertaining to transparent or reflective surfaces depicted in the target image 150. To this end, the LLM 120 may analyze the inputted target image 150 to identify lighting-related attributes LA for instance in form of elements, characteristics, portions or regions in the target image 150 pertaining to or related to lighting devices 160 depicted in the target image 150. In practice, the LLM 120 may identify lighting-related attributes LA in the target image 150 by being prompted to do so. Lighting-related attributes LA are schematically illustrated in and will be further discussed in conjunction with Figs. 3 and 4 below.

[0062] The LLM 120 of the depicted system 100 of Fig. 1 is configured to classify any identified light-related attribute LA originating from a depicted light source shining through a window, a depicted light source being reflected in a window, a depicted light source being reflected in a reflective surface, a depicted light reflection in a window, a depicted light shining through a window and / or a depicted light reflection in a reflective surface as a light interference-related attribute IA. In practice, the LLM 120 may classify identified lighting-related attributes LA in the target image 150 by being prompted to do so. In other words, the LLM 120 of Fig. 1 is configured to classify light-related attributes which are not attributable to a lighting device 160 directly depicted in the target image 150 as light interference-related attribute IA. Interference-related attributes IA are schematically illustrated in and will be further discussed in conjunction with Figs. 3 and 4 below.

[0063] To this end, the method 500, proceeds by analyzing 504, using the identification and classification system 120 (the LLM of Fig, 1 according to the present example) the target image 150 to identify lighting-related attributes LA in the target image 150, and to classify one or more lighting-related attributes LA, among the identified lighting-2024PF80406

[0064] 10

[0065] related attributes LA, as light interference-related attributes IA pertaining to transparent or reflective surfaces depicted in the target image 150.

[0066] The depicted system 100 of Fig. 1 further comprises a processing device 130. According to examples, the processing device 130 may include one or more of a CPU ("Central Processing Unit"), a DSP ("Digital Signal Processor"), a microprocessor, a microcontroller, an ASIC ("Application-Specific Integrated Circuit"), a combination of discrete analog and / or digital components, or some other programmable logical device, such as an FPGA ("Field Programmable Gate Array"). The processing device 150 may be cloud based or implemented in the cloud. The processing device 130 and the LLM 120 (the identification and classification system 120) may be jointly implemented or formed using the same hardware or partially using the same hardware.

[0067] The processing device 130 of the depicted system 100 of Fig. 1 is configured to select one or more one or more lighting devices 200 from the database 110. To this end, the processing device 150 is connected to the LLM 120 such that the processing device may communicate with the LLM 120 to e.g. receive data from the LLM 120. More specifically, the processing device 150 of Fig. 1 is configured to select one or more one or more lighting devices 200 from the database 110 based on a subset LA' of the identified lighting-related attributes LA. The subset LA' used for the selection performed by the processing device 150 comprises the identified lighting-related attributes LA except the light interference-related attributes IA. Thus, the processing device 150 is configured to receive the identified lighting-related attributes LA from the LLM 120 including the lighting-related attributes LA classified as light interference-related attributes IA. Alternatively, the processing device may receive the subset LA' from the LLM 120. In this case, the LLM 120nmay form the subset LA' from the light interference-related attributes IA and send the same to the processing device 130.

[0068] The processing device 150 of the depicted system 100 of Fig. 1 is further configured to select the one or more lighting devices 200 form the database 110 in such a way that the selected one or more lighting devices 200 having a light emission which recreates the lighting-related attributes LA of the subset LA' when used to illuminate a second space. This means in practice, the processing device 150 is configured to select one or more lighting devices 200 which when used to illuminate a second space will mimic or recrate the lighting conditions in the first space 152 depicted in the target image 150 while not taking light interference effects depicted in the target image 150 into account. Thus, the selection of the one or more lighting devices 200 made by the processing device 150 is made while disregarding lighting effects in the target image 152 pertaining to transparent or2024PF80406

[0069] 11

[0070] reflective surfaces depicted in the target image 152. In this way, non-real lighting devices, reflections, glare, lighting device or light sources being reflected or shing through a window or similar may be disregarded while the processing device 150 selecting the one or more lighting devices 200 form the database 110. The subset LA' of lighting-related attributes LA are schematically illustrated in and will be further discussed in conjunction with Figs. 3 and 4 below.

[0071] To this end, the method 500, proceeds by selecting 506, by the processing device 130, one or more lighting devices 200 from the database 110 based on the subset LA' of the identified lighting-related attributes LA. As explained above, the subset LA' comprises the identified lighting-related attributes LA except the light interference-related attributes IA. As explained above, the selected one or more lighting devices 200 are selected such that the selected one or more lighting devices 200 having a light emission which recreates the lighting-related attributes LA of the subset LA' when used to illuminate the second space.

[0072] In order to for the selected one or more lighting devices 200, i.e. the lighting devices 200 selected form the database 110, to recreate the lighting-related attributes LA of the subset LA' when used to illuminate a second space, the second space may to advantage be of the same space type as the first space 152, i.e. the space depicted in the target image. According to an example, if the target image depicts a bedroom, the selected one or more lighting devices 200 may best mimic the lighting conditions of the target image 152 when used to illuminate a bedroom.

[0073] The identification and classification system 120, i.e. the LLM 120, of the system 100 of Fig. 1 may further be configured to identify one or more ambiance-related attributes AA of the first space 152 when analyzing the target image 150. Such ambiance-related attributes AA may be indicative of a space type of the first space 152, a type of furniture present in the first space 152 or an atmosphere type in the first space 152 to give a few non-limiting examples. Ambiance-related attributes AA will be further discussed in conjunction with Figs. 3 and 4 below.

[0074] In case the LLM 120, of the system 100 of Fig. 1 is configured to identify one or more ambiance-related attributes AA of the first space 152 when analyzing the target image 150, the processing device 130 may utilize such identified ambiance-related attributes AA of the first space 152 when selecting one or more lighting devices 200. More specifically, the LLM 120, of the system 100 may further utilize such ambiance related attributes AA and select one or more lighting devices 200 from the database 110 based on the identified ambience-related attributes AA. In this way, one or more lighting devices 200 may be2024PF80406

[0075] 12

[0076] selected from the database 110 by the LLM such that a target ambience condition AA of the first space is mimicked in the second space when the selected one or more lighting devices 200 are used to illuminate the second space.

[0077] The processing device 130 of the depicted system of Fig. 1 may be further configured to classifying one or more lighting-related attributes LA, among the identified lighting-related attributes LA, as lighting device-related attributes DA pertaining to a respective lighting device 160 depicted in the target image 150. In other words, the analyzing 504 of the method 500 may associate one or more identified lighting-related attributes LA with a respective lighting device 160 depicted int the target image 150. In this way, certain lighting-related attributes LA may be classified as directly related to a lighting device 160 depicted in the target image 150. This approach enables that the one or more lighting devices 200 may be selected form the database 110 also based on the lighting device-related attributes DA. Thus, the one or more lighting devices 200 may be more accurately selected form the database 110 since the selection may take characteristics of lighting devices 160 depicted in the target image 150 into account.

[0078] To this end, a respective target feature vector representative of features of an associated lighting device-related attribute DA may be generated by the LLM 120 (the identification and classification system 120) by analyzing characteristics of the lighting device-related attribute DA. Alternatively, the processing device 130 may generate such a target feature vector from features of a lighting device-related attribute DA received by the LLM 120.

[0079] Further, to this end, in case the LLM 120, of the system 100 of Fig. 1 is configured to identify one or more ambiance-related attributes AA of the first space 152 when analyzing the target image 150, the LLM 120 may utilize such identified ambiance-related attributes AA of the first space 152 when generating target feature vectors. In this way, a respective target feature vector representative of features of an associated lighting device-related attribute DA and the ambiance-related attributes AA may be generated by the LLM 120 by analyzing characteristics of the lighting device-related attribute DA and of the ambiance-related attributes AA. Alternatively, the processing device 130 may generate such a target feature vector from features of a lighting device-related attribute DA and the ambiance-related attributes AA received by the LLM 120.

[0080] By generating target feature vectors of the above kind, the selection 506 of the one or more lighting devices 200 from the database 110 may be made utilizing the target feature vectors. In such a case, the database 110, is a vector type database 110, like the2024PF80406

[0081] 13

[0082] database 110 of the exemplifying system 100 of Fig. 1. Thus, each lighting device 200 of the database 110 is associated with a predetermined feature vector 202 as explained above. In such a case, the processing device 130 may be configured such that the selection 506 of one or more lighting devices 200 from the database 110 is done by selecting a respective lighting device 200 of the database 110 associated with a respective feature vector 202 similar to the target feature vector by performing a similarity search in the database. This means in practice, that a lighting device 200 similar to a lighting device 160 depicted in the target image may be selected 506 by the processing device 130 according to the method 500. Not only may a lighting device 200 with a similar appearance to a lighting device 160 depicted in the target image be selected (which would only require state of the art image analysis), but a lighting device 200 having a light emission which recreates relevant lighting conditions when used to illuminate a second space may be selected. Moreover, a lighting device 200 capable of mimicking an ambiance or atmosphere of the target image 152 when used to illuminate a second space may be selected.

[0083] The processing device 130 of the depicted system 100 may further be configured to generate a user comprehensible recommendation 170 of lighting devices comprising the one or more selected lighting devices 200. Thus the processing device may for instance generate text 170a or images 170b identifying the selected one or more lighting devices 200. Any type of user comprehensible recommendation 170 may be used to advantage.

[0084] In the following selected portions of the working principle of the method 500 of Fig. 2, and hence the system 100 of Fig. 1 will be described on a simplified exemplifying level while referring to Figs. 3 and 4.

[0085] Fig. 3 illustrates an example of a target image 150 depicting a space 152 in the form of a modem room with a large glass window.

[0086] Fig. 4 illustrates an example of a target image 150 depicting a space 152 in the form of a luxurious living room with a large glass window.

[0087] Now especially referring to Fig. 3. When the target image 150 of Fig. 3 has been inputted into the LLM 120 of the system of Fig. 1, the LLM 120 may, according to the method 500 of Fig. 2 analyze 504 the target image 150 to identify lighting-related attributes LA in the target image 150 as have been described above. This may, according to an example, be done by prompting the LLM 120 to do so. In the target image 150 of Fig. 3, the LLM 120 may for example identify one or more lighting-related attributes LA attributable to the glaring light 180 visible outside of the space 152 through the glass window depicted2024PF80406

[0088] 14

[0089] centrally in the target image 150. Further, lighting-related attributes LA attributable to the table lamps 181, 182 on either side of the window may further be identified. Moreover, one or more lighting-related attributes LA attributable to the reflection 183 in the floor of the depicted space 152 may further be identified. Moreover, one or more lighting-related attributes LA attributable to the cove light 184 close to the ceiling of the depicted space 152 may further be identified.

[0090] The LLM 120 may according to the method 500 of Fig. 2 analyze 504 the target image 150 to classify the identified lighting-related attributes LA of the target image 150. In the target image 150 of Fig. 3, the LLM 120 may for example classify the one or more lighting-related attributes LA attributable to the to the glaring light 180 visible outside of the space 152 and the one or more lighting-related attributes LA attributable to the reflection 183 in the floor of the depicted space 152 as light interference-related attributes IA, as have been described above. Hence, the LLM 120 may classify the identified lighting-related attributes LA, 180, 183 pertaining to transparent or reflective surfaces depicted in the target image 150 as interference-related attributes IA, 180, 183. As depicted in Fig. 3, the identified one or more lighting-related attributes LA, 180 pertain to a transparent surface, in form of the window, depicted in the target image 150, whereas the identified one or more lighting-related attributes LA, 183 pertain to a reflective surface, in form of the floor, depicted in the target image 150.

[0091] Followingly, the processing device 130 may, according to the method 500 of Fig. 2 proceed by selecting 506 one or more lighting devices 200 from the database 110 based on a subset LA' of the identified lighting-related attributes LA. As described above, the subset LA' comprises the identified lighting-related attributes LA except the light interference-related attributes IA. Thus, in the target image 150 of Fig. 3, the processing device 130 may select one or more lighting devices 200 from the database 110 based on the identified lighting-related attributes LA, 181, 182, 184 attributable to the actual lighting devices depicted in the target image 150, and hence depicted within the space 152 of the target image 150.

[0092] As have been described above, the processing device 130 may select one or more lighting devices 200 from the database 110 having a light emission which recreates the lighting-related attributes LA of the subset LA' when used to illuminate a second space. In other words, the processing device 130 may select one or more lighting devices 200 from the database 110 based on the identified lighting-related attributes LA, 181, 182, 184 attributable to the actual lighting devices 160 depicted in the target image 150 such that the selected one2024PF80406

[0093] 15

[0094] or more lighting devices 200 recreates the lighting-related attributes LA, 181, 182, 184 when used to illuminate another space than the one 152 depicted in the target image 150.

[0095] As have been described above, the LLM 120 may be further be configured to identify one or more ambiance-related attributes AA of the first space 152 when analyzing the target image 150, like the target image 150 of Fig. 3. According to an example, the LLM 120 may identify ambiance-related attributes AA in the target image 150 indicative of that the space 152 depicted in the target image of Fig. 3 is a living room. Correspondingly, the LLM 120 may identify ambiance-related attributes AA in the target image 150 indicative of that the space 152 depicted in the target image of Fig. 3 comprises a sofa and a table. Correspondingly, the LLM 120 may identify ambiance-related attributes AA in the target image 150 indicative of that the space 152 depicted in the target image of Fig. 3 has soft dim atmosphere. To this end, selecting the one or more lighting devices 200 may further be based on the identified one or more ambience-related attributes AA. In this way, also a target ambience condition of the depicted first space 152 may be mimicked in a second space, i.e. a room, when the selected one or more lighting devices 200 are used to illuminate the second space. Thus, the selection of the one or more lighting devices 200 from the database 110 may be further enhanced in the sense that the selected one or more lighting devices 200 more closely mimics the overall atmosphere including the lighting conditions of the first space 152 depicted in the target image 150.

[0096] As have been described above, the LLM 120 may be further be configured to classify one or more lighting-related attributes LA, among the identified lighting-related attributes LA, as lighting device-related attributes DA pertaining to a respective lighting device depicted in the target image 150, like the target image 150 of Fig. 3. In this way, the selection of the one or more lighting devices 200 from the database 110 may be further enhanced in the sense that the selected one or more lighting devices 200 more closely resembles the lighting devices depicted in the target image the lighting conditions of the first space 152 depicted in the target image 150.

[0097] Now especially referring to Fig. 4. When the target image 150 of Fig. 4 has been inputted into the LLM 120 of the system of Fig. 1, the LLM 120 may, according to the method 500 of Fig. 2 analyze 504 the target image 150 to identify lighting-related attributes LA in the target image 150 as have been described above. In the target image 150 of Fig.4, the LLM 120 may for example identify one or more lighting-related attributes LA attributable to the floor lamp 190. Further, one or more lighting-related attributes LA attributable to the prominent refection 191 of the floor lamp 190 in the window may be2024PF80406

[0098] 16

[0099] identified. Further, one or more lighting-related attributes LA attributable to the table lamps 192 may further be identified. Further, one or more lighting-related attributes LA attributable to the wall lamp 193 above the TV may be identified.

[0100] The LLM 120 may according to the method 500 of Fig. 2 analyze 504 the target image 150 to classify the identified lighting-related attributes LA of the target image 150. In the target image 150 of Fig. 4, the LLM 120 may, for example classify the one or more lighting-related attributes LA attributable to the reflection 191 of the floor lamp 190 as light interference-related attributes IA, 191, as have been described above.

[0101] Followingly, the processing device 130 may, according to the method 500 of Fig. 2 proceed by selecting 506 one or more lighting devices 200 from the database 110 based on a subset LA' of the identified one or more lighting-related attributes LA, where the subset LA' comprises the identified lighting-related attributes LA, 190, 192, 193 except the one or more lighting-related attributes classified as light interference-related attributes IA, 193. Thus, the processing device 130 may, according to the method 500 of Fig. 2 proceed by selecting 506 one or more lighting devices 200 from the database 110 while disregarding the reflection 191 of the floor lamp 190.

[0102] The person skilled in the art realizes that the present invention by no means is limited to the preferred embodiments described above. On the contrary, many modifications and variations are possible within the scope of the appended claims. For example, while in the above, an example embodiment of the method 500 has been disclosed with reference to the system 100, it is noted that the concept also is applicable to other types of systems 100 formed of other components or having other designs or architectures. Moreover, while in the above, the identification and classification system 120 has been described in a context where the identification and classification system 120 is an LLM , the identification and classification system 1 120 may be of any suitable type. In practice, the identification and classification system 120 may be embodied in the form of an Al model 120 or machine learning model 120, such as a neural network model 120. Moreover, the selection of the one or more lighting devices 200 from the database 110 may be made using any suitable selection technique, such as a retrieval augmented generation, RAG, based selection technique. For instance, the concept is also applicable to other environments than spaces. Thus, the target image may for instance depict an outdoor environment or a multitude of spaces.

Claims

2024PF8040617CLAIMS:

1. A method (500) of selecting one or more lighting devices (200) from a database (110) of lighting devices (200) based on a target image (150) depicting a first space (152), such as a room (152), the method (500) comprising:inputting (502) the target image (150) into an identification and classification system (120) configured to identify and classify lighting-related attributes in images,analyzing (504), using the identification and classification system (120), the target image (150) to identify lighting-related attributes (LA) in the target image (150), and to classify one or more lighting-related attributes (LA), among the identified lighting-related attributes (LA), as light interference-related attributes (IA) pertaining to transparent or reflective surfaces depicted in the target image (150),selecting (506), by a processing device (130), one or more lighting devices (200) from the database (110) based on a subset (LA') of the identified lighting-related attributes (LA), the subset (LA') comprising the identified lighting-related attributes (LA) except the light interference-related attributes (IA), wherein the selected one or more lighting devices (200) having a light emission which recreates the lighting-related attributes (LA) of the subset (LA') when used to illuminate a second space.

2. The method (500) of claim 1, wherein the classification and identification system (120) comprises an artificial intelligence, Al, model (120) trained to identify and classify lighting-related attributes in images, andwherein the analyzing (504) comprises using the Al model (120).

3. The method (500) according to claim 1 or 2, wherein an identified light-related attribute (LA) originating from one or more of; a depicted light source shining through a window, a depicted light source being reflected in a window, a depicted light source being reflected in a reflective surface, a depicted light reflection in a window, a depicted light shining through a window and a depicted light reflection in a reflective surface is classified as a light interference-related attribute (IA).2024PF80406184. The method (500) according to any one of the preceding claims, wherein the analyzing (504) further comprises:identifying one or more ambiance-related attributes (AA) of the first space (152).

5. The method (500) according to claim 4, wherein an identified ambience-related attribute (AA) is indicative of, a space type of the first space (152), a type of furniture present in the first space (152) or an atmosphere type in the first space (152).

6. The method (500) according to claim 4 or 5, wherein the selecting (506) further comprises:selecting the one or more lighting devices (200) based on the identified one or more ambience-related attributes (AA), such that a target ambience condition of the first space (152) is mimicked in the second space when the selected one or more lighting devices (200) are used to illuminate the second space.

7. The method (500) according to any one of the preceding claims, wherein the first space (152) is of a first space type and wherein the second space is of the first space type.

8. The method (500) according to any one of the preceding claims, wherein the analyzing (504) further comprises:classifying one or more lighting-related attributes (LA), among the identified lighting-related attributes (LA), as lighting device-related attributes (DA) pertaining to a respective lighting device (160) depicted in the target image (150).

9. The method (500) according to claim 8, wherein the selecting (506) further comprises:selecting the one or more lighting devices (200) based on the lighting devicerelated attributes (DA).

10. The method (500) according to claim 8, further comprising:generating, for each lighting device-related attribute (DA), a target feature2024PF8040619vector representative of features of the lighting device-related attribute (DA) by analyzing characteristics of the lighting device-related attribute (DA).

11. The method (500) according to claim 8 when dependent on claim 4, further comprising:generating, for each lighting device-related attribute (DA), a target feature vector representative of features of the lighting device-related attribute (DA) and the ambiance-related attributes (AA) by analyzing characteristics of the lighting device-related attribute (DA) and of the ambiance-related attributes (AA).

12. The method according to claim 10 or 11, wherein the database (110) is a vector type database (110), wherein each lighting device (200) of the database (110) is associated with a predetermined feature vector (202), andwherein the selecting (506) one or more lighting devices (200) comprises selecting a lighting device (200) associated with a feature vector (202) similar to the target feature vector by performing a similarity search in the database (110).

13. The method (500) according to any one of the preceding claims, further comprising:generating (508), by the processing device (130), a user comprehensible recommendation (170) of lighting devices (200) comprising the one or more selected lighting devices (200).

14. The method (500) according to claim 2, wherein the Al model (120) is a neural network model (120), or a large language model, LLM (120).

15. A computer program product comprising computer program code means which, when executed on a computing device having a processing system, cause the processing system to perform all of the steps of the method according to any one of claims 1 to 14.

Images