Method for distinguishing user feedback to image

Abstract

A method for distinguishing user feedback to an image is disclosed. The method includes providing an image of a scene including an environment, a luminaire design of the luminaire, and a light effect of the luminaire in the environment, analyzing the image, determining a first saliency value of the luminaire design in the image, determining a second saliency value of the light effect in the image, where the second saliency value is different than the first saliency value. The method further includes receiving user feedback on the image and associating the user feedback with the luminaire design and / or the light effect based on the first and second saliency values.

Description

Technical Field

[0001] The present invention relates to a method for distinguishing user feedback on an image. The present invention also relates to a computer program and a controller for distinguishing user feedback on an image. Background Art

[0002] As the availability of luminaire designs increases, so does the challenge of identifying and providing personalized recommendations for users. In recent years, such personalization has focused on selecting luminaire designs that match the user's aesthetic tastes and desires. As an example, in patent application WO 2014064634, a method for assisting a user in selecting a lighting device based on a scene and light effects selected by the user is proposed. Patent application WO 2014087274 relates to assisting a user in selecting a lighting device design by receiving an image of a scene and analyzing the image in order to select or generate a lighting device design. The light effects that are part of the scene (e.g., a wide beam downlight from the center of the ceiling applied to a living room) may be first presented to the user, allowing the user to select the light effect before selecting the lighting device design. Summary of the invention

[0003] The inventors have appreciated that when a user is presented with an image of a luminaire design of a luminaire and its effect in an environment, the user may evaluate the image as attractive based on a combination of both the luminaire design (e.g. luminaire shape, texture, material of the luminaire housing, etc.) and the light effect applied by the luminaire in the environment (e.g. applied spatial spectral distribution, temporal dynamics of the effect, light intensity, beam shape, etc.). However, the user may not be able to verbally explain the reasoning behind such a preference. Thus, the user's preference (appreciation or dislike feedback) for an image may be directed to / associated with the luminaire design, the light effect of the luminaire design on the environment, or a combination of both. For example, a user's rejection of an image of a scene including an environment, a luminaire design of a luminaire and its light effect may be due to the user appreciating the luminaire design (or a particular aspect of the luminaire design) but disliking the light effect of the luminaire on the environment depicted in the image. Similarly, a user's rejection of an image may be due to the user appreciating the light effect but disliking the material of the luminaire, etc.

[0004] One purpose is to improve learning of user preferences.

[0005] According to a first aspect, the object is achieved by a method for distinguishing user feedback on an image, the method comprising the following steps:

[0006] - providing an image of a scene including an environment, a luminaire design of a luminaire and light effects of the luminaire in the environment;

[0007] - Analyze images;

[0008] - determining a first salient value of the illuminator design in the image;

[0009] - determining a second significance value of the light effect in the image;

[0010] - Receive user feedback on images;

[0011] - correlating the user feedback with the luminaire design and / or the light effect based on the first and second saliency values.

[0012] An image of a scene including an environment, a luminaire design of a luminaire, and a light effect of the luminaire in the environment is analyzed to determine a first saliency value (saliency level) of the luminaire design and a second saliency value (saliency level) of the light effect of the luminaire. The second saliency value may be different from the first saliency value. In the context of the present disclosure, a saliency value may refer to a value indicating the noticeability (importance / prominence) of a luminaire design or a light effect in an image, respectively. For example, a first image of a luminaire design in which the light effect of the luminaire design is negligible (e.g., an image of the luminaire design in which only ambient lighting is present during daytime) has a low light effect second saliency value. In contrast, a second image of the same luminaire design in the same environment but during nighttime in which the light effect of the luminaire design is prominent (highly visible) has a high light effect second saliency value.

[0013] By providing an image of a scene comprising an environment, a luminaire design of a luminaire and a light effect of the luminaire in the environment to a user, analyzing the image to determine first and second saliency levels of the luminaire design and the light effect, respectively, the user's feedback can be associated with the luminaire design and / or the light effect based on the first and second saliency values. Thus, the method allows distinguishing whether the user's feedback is for the luminaire design or for the light effect without explicitly asking the user (who may also not be able to explain the reasoning behind the like / dislike feedback). This is beneficial because it improves learning of user preferences.

[0014] The step of associating the user feedback may comprise associating the user feedback with the light effect if the second saliency value is greater than the first saliency value, and associating the user feedback with the luminaire design if the first saliency value is greater than the second saliency value. This provides a simple way to determine whether the feedback is for the luminaire design or the light effect.

[0015] The step of associating user feedback may include associating user feedback with luminaire design as a function of a first significance level, and associating user feedback with light effects as a function of a second significance value. In some cases, the user's feedback may be related to both luminaire design and light effects. In such cases (but not limited to these cases), the method may include associating user feedback with luminaire design as a function of a first significance level, and associating user feedback with light effects as a function of a second significance value. For example, the method may include assigning a likelihood to the association of feedback with luminaire design and with light effects. The likelihood may include a probability or weight assigned to the feedback of the luminaire design / light effect. The likelihood may include a relative weight or relative probability. The likelihood may be based on a function of the significance value, such that the likelihood of the feedback being associated with the luminaire design / light effect is proportional to (similar to) the light effect significance value of the luminaire design / light effect, respectively.

[0016] The method may also include generating a first text description of the user's preference for the luminaire design and a second text description of the user's preference for the light effect using a machine learning model based on the associated feedback, and outputting the generated first and second text descriptions. The generated first and second text descriptions may be output to a user interface for presentation to the user, or, for example, as input to another machine learning model. The machine learning model may have been trained using labeled image instances with associated user feedback as input. The machine learning model may, for example, be a hierarchical combination of a convolutional neural network (CNN) responsible for image feature extraction and a long short-term memory model (LSTM) for generating text descriptions. An example of such a machine learning model may be a generative pre-trained transformer 3 (GPT-3). In this way, the user can understand his / her personal preferences for different aspects of the luminaire (i.e., the luminaire design and the light effect of the luminaire in the environment). Optionally, the user can provide feedback on the preferred text description (whether he / she thinks the preferred text description is accurate, etc.). In addition, the text description may be fed into a database including feedback from many different users. Based on statistical analysis of the text description, the system may automatically generate different user types that share similar design preferences. During reasoning, these user types may be utilized to accelerate the convergence of new users to a satisfactory luminaire lighting design.

[0017] The second saliency value may be determined based on the spread of the light effect in the environment. The image may be analyzed to determine the spread level (spatial distribution) of the light in the environment. The visual saliency level (visual saliency value) of the light effect may depend on the spread level, i.e. the spatial distribution of the light effect in the image. In other words, the second visual saliency value may depend on the extent to which the light effect of the luminaire affects the surroundings of the luminaire design in the image. The image may be analyzed to determine the number of luminaires in the image. The second saliency value may also depend on whether there is only a single luminaire visible in the image, or whether there are more than one luminaire present that also generate their own light effects in the image. Similarly, the first saliency value may depend on the spatial distribution of the luminaire design in the image. In other words, how much space the luminaire design occupies in the image.

[0018] The second saliency value may be determined based on characteristics of the environment. For example, the image may be analyzed to determine the distribution of light effects in the image. For example, an abstract environment with a uniform light effect distribution (light effects are uniformly spread around the luminaire design in the image) may have a lower second saliency value than a detailed environment (i.e. an environment with multiple elements, where the light effects are combined with the elements of the environment). The first and second saliency values ​​may also depend on the saturation of the luminaire design / light effect in the image, respectively. The first and second saliency values ​​may also depend on whether there is only a single visible luminaire in the image, or whether there are more than one luminaire present that also generate their own light effects in the image. If multiple luminaires are present in the image, the saliency of the lighting effect is more prominent.

[0019] The user's feedback may include text input. For example, the text input from the user may be received via a user interface. As a specific example, the user interface may be implemented by way of one or more web pages displayed by a user device via a web browser software program. Additionally, and / or alternatively, the user feedback may also be in the form of voice commands.

[0020] In one example, user feedback may be input data indicating a physiological change in the user (indicating appreciation or dislike of an image). For example, the user's heart rate and / or breathing rate may be received, and the user feedback may be determined based on the measured changes in the user's heart rate and / or breathing rate. In an advanced example, the user's EEG (electroencephalogram), EOG (electrooculogram), EDA (electrodermal activity), PPG (photoplethysmography), EMG (electromyographic activity), etc. may be received, for example, through an augmented reality headset, a wristband, etc., and the user feedback may be determined based on changes in these measurements of the user.

[0021] The user's feedback may also include gestures. For example, gesture input (such as finger sliding or hand movement) may be received from the user as sensed using, for example, an optical or capacitive sensor.

[0022] In some cases, it will be desirable to provide (recommend) images of luminaire designs outside of an existing list or catalog of luminaires to a user. The method may also include generating a composite image of a scene using a generative AI machine learning model, the composite image of the scene comprising additional (synthesized) luminaire designs of the additional luminaires and light effects based on associated feedback. In one embodiment, the generative AI machine learning model is a text-conditional generative adversarial network that is adjusted to generate a composite image based on a generated text description of user preferences for luminaire designs and associated light effects. In another example, a text-to-image diffusion model (such as Imagen, DALL-E 2, etc.) can be used to generate a composite image using a text description as input. Thus, a method is provided for automatically designing and generating new luminaire designs (with corresponding light effects) that better match user tastes and preferences.

[0023] The method may further include:

[0024] - Analyze the composite image;

[0025] - determining a (further) first saliency value of the illuminator design in the synthetic image;

[0026] - determining a (further) second saliency value of the light effect in the composite image;

[0027] - receiving user feedback on the synthesized image;

[0028] - correlating the user feedback with the luminaire design and / or the light effect based on the first and second saliency values.

[0029] This is beneficial because the user feedback association can in turn trigger a re-iteration of the previous steps of providing further composite images. This allows providing further composite images that are closer to the user's tastes and preferences (accelerating convergence to the user's preferred luminaire design).

[0030] The method may also include generating specifications for additional luminaire designs and specifications for light effects, and outputting these specifications to a system or service for generating (or manufacturing) additional luminaire designs. For example, the specifications for light effects may include the shape of the light effect, the pattern of the light effect, one or more colors of the light effect, and / or the location of light effect features. The specifications for luminaire designs may include the shape of the luminaire design, the size of the luminaire design, the number of lumens based on the light effect, the number of emitters and the type of emitters based on the number of lumens, and the number of drivers and the type of drivers based on the number and type of emitters. After outputting the system specifications for generating additional luminaire designs, the user may order additional (synthesized) luminaire designs, or print the designs via a 3D printer.

[0031] According to a second aspect, the object is achieved by a computer program product for a computing device, the computer program product comprising computer program code to perform any of the above methods when the computer program product is run on a processing unit of the computing device. Such a computer program product can be executed on a computer (such as a personal computer or a laptop) or a smartphone or other computing device.

[0032] According to a third aspect, the object is achieved by a controller for distinguishing user feedback on an image, the controller being configured to:

[0033] -provide an image of a scene including the environment, the luminaire design and the light effects of the luminaire in the environment;

[0034] - Analyze images;

[0035] - determining a first salient value of the illuminator design in the image;

[0036] - determining a second saliency value for the light effect in the image, wherein the second saliency value is different from the first saliency value,

[0037] - Receive user feedback on images;

[0038] - correlating the user feedback with the luminaire design and / or the light effect based on the first and second saliency values.

[0039] It should be understood that the computer program product and the controller may have similar and / or the same embodiments and advantages as the above-described method. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The above and additional objects, features and advantages of the disclosed systems, apparatus and methods will be better understood through the following illustrative and non-limiting detailed description of embodiments of the apparatus and methods with reference to the accompanying drawings, in which:

[0041] Figure 1 schematically showing an example of a scene image including an environment, a luminaire design of a luminaire and a light effect of the luminaire in the environment;

[0042] Figure 2 Schematically shown is a controller configured to provide a user with an image of a scene including an environment, a luminaire design of a luminaire and a light effect of the luminaire in the environment;

[0043] Figure 3 Schematically shown is a flow chart illustrating an embodiment of a method for distinguishing user feedback on an image;

[0044] Figure 4 An example of a scene image including an environment, a luminaire design of a luminaire and the light effect of the luminaire in the environment is schematically shown.

[0045] All the figures are schematic, not necessarily to scale, and generally show only parts which are necessary in order to elucidate the invention, wherein other parts may be omitted or merely suggested. DETAILED DESCRIPTION

[0046] Figure 1 An example of an image 100 of a scene including an environment 142, a luminaire design 102 of a luminaire, and a light effect 112 of the luminaire in the environment is shown. The environment 102 refers to the surrounding environment of the luminaire design. The environment 102 can be any type of home environment (e.g., kitchen, living room, bathroom, etc.), commercial environment (e.g., factory, restaurant, office, etc.), or solid color background environment (e.g., solid white or other color fonts).

[0047] Luminaire design 102 includes at least one light source or lamp (not shown), such as an LED-based lamp, a gas discharge lamp, or a filament bulb, etc., optionally with an associated support, housing, or other such housing. Luminaire design 102 can take any of a variety of forms, such as ceiling mounted luminaires, wall mounted luminaires, wall washers, freestanding luminaires, LED light strips, LED bulbs, laser lighting fixtures, and ultra-thin OLED luminaires, etc., and any size, shape, material, or color. In this exemplary figure, luminaire design 102 is a ceiling luminaire. Image 100 can include any number of luminaires.

[0048] The light effects 112 of the luminaire in the environment refer to the light output of at least one light source or lamp and how the light output affects the surroundings, environment, of the luminaire. The light effects 112 may include the color or color temperature of the light source, the illuminance intensity (brightness), the beam width, the beam direction, and other parameters of one or more light sources of the luminaire design 102. The image may be part of a video and the light effects 112 may include dynamic light scenes, wherein the dynamic light scenes may include light effects that change over time.

[0049] Figure 2 Schematically and exemplarily, a controller 210 is shown which is configured to provide an image 200 (eg, Figure 1200). The controller 210 may be implemented in a device, such as a desktop computer or a portable terminal (such as a laptop, tablet computer, or smartphone). The controller 210 may alternatively be implemented in the cloud, for example as a server accessible via the Internet. The image 200 is provided to the user 220, for example, via a user interface on a device such as a laptop, tablet computer, or smartphone 236. Alternatively, the image 200 may be shown to the user 220 as a video. Alternatively, the image 200 may be shown to the user 220 via an AR / VR head mounted viewer. The controller 210 is configured to analyze the image 200 and determine a first saliency value of the illuminator design in the image and a second saliency value of the light effect in the image based on the analysis. The controller 210 may be configured to apply image analysis techniques to identify the illuminator (and its design) and the light effect of the illuminator in the image 200. Image analysis techniques for identifying objects and features in images are known in the art and therefore will not be discussed in detail.

[0050] For example, the second saliency value may be determined based on the spread of the light effects 112 in the environment 142. That is, the second saliency value may be proportional to the spatial distribution of the light effects 112 in the image 100. In other words, the second visual saliency value may depend on how much space, such as the number of pixels, the light effects 112 of the luminaire occupy in the image 100. Similarly, the first saliency value may depend on the spatial distribution of the luminaire design 102 in the image 100. In other words, the first saliency value may depend on how much space, such as the number of pixels, the luminaire design 102 occupies in the image 100. In the exemplary Figure 1 , the first (visual) saliency value is higher than the second (visual) saliency value because the luminaire design 102 occupies a large portion of the image 100. The second saliency value may be further analyzed to include saliency values ​​of design aspects of the luminaire design. For example, saliency values ​​of the shape, material, style, etc. of the luminaire design 102.

[0051] Figure 4 An example of an image 400 of a scene including an environment 442, a luminaire design 402 for a luminaire, and a light effect 412 of the luminaire in the environment is shown. Figure 44, the first (visual) saliency value of the luminaire design is lower than the second (visual) saliency value of the light effect, because the light effect 412 occupies more space (higher number of pixels) in the image 400 than the luminaire design 402. The second saliency value may be determined based on characteristics of the environment. For example, an abstract environment with a uniform light effect distribution (light effects are uniformly spread around the luminaire design in the image, such as the environment 142 in the image 100) may have a lower second saliency value compared to a detailed environment (i.e. an environment with multiple elements, where the light effect is combined with elements of the environment such as, for example, the environment 442 in the image 400). The first and second saliency values ​​may also depend on the saturation of the luminaire design / light effect, respectively, in the image.

[0052] The second saliency value may also depend on the color contrast of the light effect. For example, an image with a high light effect color contrast may have a higher light effect saliency value than an image where the color (color temperature) of the light effect is homogeneous.

[0053] Additionally and or alternatively, a saliency algorithm for image saliency detection, such as the GrabCut algorithm, may be used for automatic extraction of the first and second saliency values ​​in the image.

[0054] Reference Figure 1 , receiving feedback 132 on the image 200 from the user 220. The feedback 132 may include a user rating scale, such as a numeric rating scale (e.g., a 1-10 rating scale), a binary rating scale (the user rates the image positively or negatively), a verbal rating scale, activating at least one actuator, such as a like / dislike button on the user's mobile device 236 to indicate his / her preference, the user may move his / her finger (swipe) across the screen to indicate positive / negative feedback depending on the direction of movement, etc. Additionally and / or alternatively, the feedback 132 may include input data indicative of physiological changes of the user 220. For example, the heart rate or breathing rate of the user 220 may be received by the controller 210, and the user feedback 132 may be determined based on measured changes in the heart rate, sweat rate, or breathing rate of the user 220. In an advanced example, the EEG of the user 220 may be received by the controller 210, and the user feedback 132 may be determined based on changes in the EEG measurements of the user 220.

[0055] The controller 210 may be configured to associate the feedback 132 with the light effect 112 based on the condition that if the first significance value of the light effect is greater than the second significance value of the luminaire design. Similarly, if the first significance value of the luminaire design is greater than the second significance value of the light effect, the feedback 132 is associated with the luminaire design 102. Alternatively, the controller 210 may be configured to associate the user feedback 132 with the luminaire design 102 as a function of the first significance level and to associate the user feedback 132 with the light effect 112 as a function of the second significance value. For example, a likelihood may be assigned to the association of the feedback with the luminaire design and with the light effect; wherein the likelihood of the feedback 132 being associated with the light effect (luminaire design) is proportional to the significance value of the light effect (respectively to the significance value of the luminaire design). The controller may further comprise a memory 222, which may be arranged for storing, for example, the user's feedback.

[0056] The controller 210 may also optionally be configured to generate a first text description of the user's preference for the luminaire design 102 and a second text description of the user's preference for the light effect 112 using a machine learning model based on the associated feedback 132. In one example, the first text description may be in the form of "oval suspension luminaire design". In another example, the second text description may be in the form of "homogeneously distributed blue light effect", etc. The machine learning model may have been trained using labeled image instances with associated user feedback as input. Computer vision machine learning models such as convolutional neural networks may be used to identify features in the image, such as the shape of the luminaire design, etc., while natural language processing such as recurrent neural networks like LSTMs may be used to generate text descriptions of the images that the user finds attractive. The generated text description may be output to the user 220, for example, on the user's mobile device 236, for example, via a display, an AR / VR head-mounted viewer, or a voice interface.

[0057] Figure 3 Schematically and exemplarily, a flow chart illustrating an embodiment of a method 300 for distinguishing user feedback on an image is shown, the method comprising the following steps:

[0058] - providing 302 by the controller 210 an image 100 of a scene comprising an environment 142 , a luminaire design 102 of a luminaire and a light effect 112 of the luminaire in the environment;

[0059] - analyzing 304 the image 100 by the controller 210 ;

[0060] - determining 306, by the controller 210, a first saliency value of the luminaire design in the image;

[0061] - determining 308, by the controller 210, a second saliency value of the light effect in the image, wherein the second saliency value is different from the first saliency value,

[0062] - receiving 310 user feedback 132 on the image 100 by the controller 210, and

[0063] Associating 312 , by the controller 210 , the user feedback 132 with the luminaire design 102 and / or the light effect 112 based on the first and second saliency values.

[0064] The method 300 may include generating 314 a synthetic image of the scene including additional illuminator designs and light effects based on the associated feedback using a generative AI machine learning model. For example, a text-conditional generative adversarial network (e.g., a TAC-GAN model) may be used to synthesize an image from a text description by conditioning the generated synthetic image according to the text description. In another example, a text-to-image diffusion model (such as Imagen, DALL-E 2, etc.) may be used to generate a synthetic image using a text description as input.

[0065] Such generative AI (language) models are known in the prior art and will therefore not be discussed further.

[0066] Method 300 may also include repeating steps 302 to 312 for the composite image.

[0067] The method 300 may also include generating 316 specifications for the additional luminaire design and specifications for the light effect, and outputting 318 these specifications to a system or service for generating (or manufacturing) the additional luminaire design. After outputting the system specifications for generating the additional luminaire design, the user may order the additional (synthesized) luminaire design or print the design via a 3D printer.

[0068] The method 300 may be performed by computer program code of a computer program product when the computer program product is executed on a processing unit of a computing device, such as the controller 210 .

[0069] It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design many alternative embodiments without departing from the scope of the appended claims.

[0070] In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The use of the verb "comprise" and its conjugations does not exclude the presence of elements or steps other than those stated in a claim. The article "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several distinct elements and by means of a suitably programmed computer or processing unit. In a device claim enumerating several means, several of these means may be implemented by one and the same item of hardware. The mere fact that certain measures are recited in mutually different dependent claims does not mean that a combination of these measures cannot be used to advantage.

[0071] Aspects of the present invention may be implemented in a computer program product, which may be a collection of computer program instructions stored on a computer-readable storage device, which may be executed by a computer. The instructions of the present invention may be any interpretable or executable code mechanism, including but not limited to scripts, interpretable programs, dynamic link libraries (DLLs), or Java classes. Instructions may be provided as complete executable programs, partially executable programs, as modifications (e.g., updates) to existing programs, or extensions (e.g., plug-ins) to existing programs. In addition, portions of the present invention may be distributed across multiple computers or processors, even on the "cloud".

[0072] Storage media suitable for storing computer program instructions include all forms of non-volatile memory, including but not limited to EPROM, EEPROM and flash memory devices, magnetic disks such as internal and external hard drives, removable disks and CD-ROM disks. The computer program product may be distributed on such storage media or may be provided for download via HTTP, FTP, email or through a server connected to a network such as the Internet.

Claims

1. A method for distinguishing whether user feedback (132) on an image (100) is associated with a luminaire design and / or a light effect of the luminaire, the method comprising the following steps: - providing (302) an image of a scene, the image comprising an environment (142), a luminaire, the luminaire having a luminaire design (102) and a light effect (112) of the luminaire in the environment; - analyzing (304) the image to (i) determine (306) a first saliency value of the luminaire design in the image, and (ii) determine (308) a second saliency value of the light effect in the image, wherein the second saliency value is different from the first saliency value; - receiving (310) user feedback on the image; - Associating the user feedback with the luminaire design and / or the light effect based on the first and second saliency values ​​(312).

2. The method according to claim 1, wherein the method further comprises: - generating, using a machine learning model, a first textual description of the user's preference for a luminaire design and a second textual description of the user's preference for a light effect based on the associated user feedback; - outputting the generated first and second text descriptions.

3. The method according to claim 1 or 2, wherein the step of associating user feedback comprises associating the user feedback with the light effect if the second saliency value is greater than the first saliency value, and associating the user feedback with the luminaire design if the first saliency value is greater than the second saliency value.

4. The method according to claim 1 or 2, wherein the step of associating user feedback comprises associating user feedback with luminaire design as a function of a first significance level and associating user feedback with light effects as a function of a second significance value.

5. A method according to claim 1, wherein the step of analyzing the image includes analyzing the image to determine the spatial distribution of light effects and the spatial distribution of illuminator designs in the image, and wherein the step of determining the first and second saliency values ​​includes determining the first and second saliency values ​​based on the illuminator design and the spatial distribution of light effects in the environment, respectively.

6. The method of claim 1, wherein the step of analyzing the image comprises analyzing the image to determine a characteristic of the environment, and wherein the step of determining the first and second saliency values ​​comprises determining the first and second saliency values ​​based on the characteristic of the environment.

7. A method according to any one of the preceding claims, wherein the step of receiving user feedback comprises receiving text and / or voice input from the user.

8. A method according to any preceding claim, wherein the step of receiving user feedback comprises receiving input indicative of a physiological change in the user.

9. A method according to any preceding claim, wherein the step of receiving user feedback comprises receiving gesture input from a user.

10. A method according to any one of the preceding claims, wherein the method further comprises generating a composite image of the scene using a generative AI machine learning model and based on associated user feedback, the composite image comprising a further luminaire design of the further luminaire and the light effects of the further luminaire.

11. A method according to claim 10 when dependent on claim 2, wherein the generative AI machine learning model is a text-conditional generative adversarial network or a text-to-image diffusion model, which is adjusted to generate a synthetic image based on the generated text description.

12. The method according to claim 10 or 11, wherein the method further comprises: - analyzing the composite image to (i) determine a further first significance value of the luminaire design in the composite image, and (ii) determine a further second significance value of the light effect in the composite image; - receiving further user feedback on the composite image; - associating further user feedback with the luminaire design and / or the light effect based on the further first and second saliency values.

13. The method according to claim 10 or 11, wherein the method further comprises: - generating a specification for a further luminaire design and exporting the specification to a system or service for generating (or manufacturing) the further luminaire design.

14. A computer program product for a computing device, the computer program product comprising computer program code to perform the method of any one of claims 1 to 13 when the computer program product is run on a processing unit of the computing device.

15. A controller (210) for distinguishing whether user feedback on an image is associated with a luminaire design and / or a light effect of the luminaire, the controller being configured to: - providing an image of a scene, the image comprising an environment, a luminaire having a luminaire design and light effects of the luminaire in the environment; - analyzing the image to (i) determine a first significance value of the luminaire design in the image, and (ii) determine a second significance value of the light effect in the image; Where the second significant value is different from the first significant value; - Receive user feedback on images; - correlating the user feedback with the luminaire design and / or the light effect based on the first and second saliency values.

Images