Method and system for determining naturalness of artificial intelligence based images

Abstract

Disclosed subject matter relates to a method and naturalness determination system for determining naturalness of AI based images. The method includes receiving an AI based facial image of a human, comprising one or more facial features extracted from the AI based facial image. The method further includes obtaining a feature confidence score for each of the one or more facial features using feature-specific naturalness evaluation models and further generating a facial confidence score for the AI based facial image using a face-specific naturalness evaluation model. Thereafter, the method includes obtaining a naturalness confidence score for the AI based facial image based on a correlation of the feature confidence score of each of the one or more facial features and the facial confidence score. Finally, the method includes determining naturalness of the AI based facial image by classifying the AI based facial image as natural or un-natural.

Description

METHOD AND SYSTEM FOR DETERMINING NATURALNESS OF ARTIFICIAL INTELLIGENCE BASED IMAGES

[0001] The present subject matter relates generally to a field of image processing. Particularly, but not exclusively the disclosure relates to a method and a system for determining naturalness of Artificial Intelligence (AI) based images.

[0002] In recent years, due to the rapid development of computer vision technology and deep learning technology, editing and synthesising face images has become easier. Accordingly, in various image processing techniques, images are artificially generated using Deep Neural Network (DNN) models and Artificial Intelligence (AI) models. The AI based images are vivid due to the current progress in image processing technologies. Therefore, it is challenging to accurately judge an authenticity of the AI based images.

[0003] Existing image evaluation technologies are focused on distinguishing a real face image from a forged face image which may be generated using the AI models. However, the existing image evaluation technologies may not always be reliable for images generated by various means. For example, in a scenario where a user may wish to enhance an old image using AI techniques, there is a possibility that the corresponding enhanced image may not be satisfactory to the user. Subsequently, the user may further wish to check naturalness of the enhanced image. In this scenario, the existing image evaluation technologies may always consider the enhanced image as unnatural since they are generated based on the AI techniques.

[0004] Further, image evaluator models in the existing image evaluation technologies are trained to predict the naturalness of the face images. The existing image evaluator models tend to struggle to address all artifacts in the face images, as the models are trained to generate a single confidence score for a complete face image. Therefore, evaluation results generated by the existing image evaluator models are not reliable, as minor artifacts such as mouth and eyes, may miss out while performing computation of the confidence score.

[0005] Thus, the existing image evaluation technologies are not reliable for distinguishing natural and unnatural images, which are generated based on the AI techniques.

[0006] The information disclosed in this background of the disclosure section is only for enhancement of understanding of the general background of the disclosure and should not be taken as an acknowledgement or any form of suggestion that this information forms prior art already known to a person skilled in the art.

[0007] One or more shortcomings of the prior art may be overcome, and additional advantages may be provided through the present disclosure. Additional features and advantages may be realized through the techniques of the present disclosure. Other embodiments and aspects of the disclosure are described in detail herein and are considered a part of the claimed disclosure.

[0008] According to an embodiment of the disclosure, a method of determining naturalness of Artificial Intelligence (AI) based images is disclosed. The method includes receiving an AI based facial image of a human from predefined sources, the AI based facial image comprising one or more facial features extracted from the AI based facial image. The method includes obtaining a feature confidence score and a facial confidence score. The feature confidence score is generated for each of the one or more facial features using feature-specific naturalness evaluation models trained for evaluating the respective facial features. The facial confidence score is generated for the AI based facial image using a face-specific naturalness evaluation model trained for evaluating a human face. Thereafter, the method includes obtaining a naturalness confidence score for the AI based facial image based on a correlation of, the feature confidence score for each of the one or more facial features and the facial confidence score. Finally, the method includes determining naturalness of the AI based facial image by classifying the AI based facial image based on the naturalness confidence score.

[0009] According to an embodiment of the disclosure, a naturalness determination system for determining naturalness of Artificial Intelligence (AI) generated images is disclosed. The naturalness determination system comprises a processor and a memory communicatively coupled to the processor, where the memory stores processor executable instructions, which, on execution, may cause the naturalness determination system to receive an AI based facial image of a human, generated using AI, the AI based facial image comprising one or more facial features extracted from the AI based facial image. The naturalness determination system further obtains a feature confidence score and a facial confidence score. The feature confidence score is generated for each of the one or more facial features using feature-specific naturalness evaluation models trained for evaluating the respective facial features. The facial confidence score is generated for the AI based facial image using a face-specific naturalness evaluation model trained for evaluating a human face. Thereafter, the naturalness determination system obtains a naturalness confidence score for the AI based facial image based on a correlation of, the feature confidence score for each of the one or more facial features and the facial confidence score. Finally, the naturalness determination system determines naturalness of the AI based facial image by classifying the AI based facial image based on the naturalness confidence score.

[0010] According to an embodiment of the present disclosure, a computer-readable storage medium storing instructions is provided. The instructions, when executed by at least one processor, may cause the at least one processor to receive an AI based facial image of a human, generated using AI, the AI based facial image comprising one or more facial features extracted from the AI based facial image. The instructions, when executed by the at least one processor, may cause the at least one processor to obtain a feature confidence score and a facial confidence score. The feature confidence score is generated for each of the one or more facial features using feature-specific naturalness evaluation models trained for evaluating the respective facial features. The feature confidence score is generated for each of the one or more facial features using feature-specific naturalness evaluation models trained for evaluating the respective facial features. The facial confidence score is generated for the AI based facial image using a face-specific naturalness evaluation model trained for evaluating a human face. The instructions, when executed by at least one processor, may cause the at least one processor to obtain a naturalness confidence score for the AI based facial image based on a correlation of the feature confidence score for each of the one or more facial features and the facial confidence score. The instructions, when executed by at least one processor, may cause the at least one processor to determine naturalness of the AI based facial image by classifying the AI based facial image based on the naturalness confidence score.

[0011] The foregoing summary is illustrative only and is not intended to be in any way limiting. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features will become apparent by reference to the drawings and the following detailed description.

[0012] The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate exemplary embodiments and, together with the description, serve to explain the disclosed principles. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The same numbers are used throughout the figures to reference like features and components. Some embodiments of system and / or methods in accordance with embodiments of the present subject matter are now described, by way of example only, and with reference to the accompanying figures, in which:

[0013] Figure 1 shows an embodiment of the present disclosure for determining naturalness of Artificial Intelligence (AI) based images;

[0014] Figure 2 shows a detailed block diagram of naturalness determination system in accordance with an embodiment of the present disclosure;

[0015] Figure 3A and 3B show exemplary illustrations of determining naturalness of Artificial Intelligence (AI) based images using naturalness evaluation models and correlation models in accordance with an embodiment of the present disclosure;

[0016] Figure 4A shows exemplary Artificial Intelligence (AI) based images classified as natural in accordance with an embodiment of the present disclosure;

[0017] Figure 4B shows exemplary Artificial Intelligence (AI) based images classified as un-natural in accordance with an embodiment of the present disclosure;

[0018] Figure 5A, 5B, and 5C show exemplary determination of naturalness of AI based images using correlation models in accordance with an embodiment of the present disclosure;

[0019] Figure 6 is a flowchart illustrating a method for naturalness determination system for determining naturalness of Artificial Intelligence (AI) based images in accordance with an embodiment of the present disclosure; and

[0020] FIGURE 7 is a block diagram of an exemplary system for implementing an embodiment of the present disclosure.

[0021] It should be appreciated by those skilled in the art that any block diagrams herein represent conceptual views of illustrative systems embodying the principles of the present subject matter. Similarly, it will be appreciated that any flow charts, flow diagrams, state transition diagrams, pseudo code, and the like represent various processes which may be substantially represented in computer readable medium and executed by a computer or processor, whether or not such computer or processor is explicitly shown.

[0022] In the present document, the word "exemplary" is used herein to mean "serving as an example, instance, or illustration." Any embodiment or implementation of the present subject matter described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments.

[0023] While the disclosure is susceptible to various modifications and alternative forms, specific embodiment thereof has been shown by way of example in the drawings and will be described in detail below. It should be understood, however, that it is not intended to limit the disclosure to the forms disclosed, but on the contrary, the disclosure is to cover all modifications, equivalents, and alternative falling within the scope of the disclosure.

[0024] The terms “comprises,” “comprising,” “includes” or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a setup, device, or method that includes a list of components or steps does not include only those components or steps but may include other components or steps not expressly listed or inherent to such setup or device or method. In other words, one or more elements in a system or apparatus proceeded by “comprises… a” does not, without more constraints, preclude the existence of other elements or additional elements in the system or method.

[0025] Disclosed herein is a method and a system for determining naturalness of Artificial Intelligence (AI) based images. Existing image evaluation technologies are not reliable for determining naturalness of AI based images as all AI generated images are classified as unnatural. Therefore, this leads to misclassification of the AI generated images thereby leading to dissatisfaction to end users.

[0026] Therefore, to solve the above problem, the present disclosure discloses a method and a system for determining naturalness of Artificial Intelligence (AI) based images. The present disclosure facilitates determination of the naturalness of the AI based images based on confidence scores generated for each feature in the AI based image. Particularly, the confidence scores are generated by individual feature specific models, that are trained to intelligently merge confidence score from various facial features into a single comprehensive confidence score. The resulting comprehensive confidence score after amalgamating all the confidence scores is significantly more reliable due to model ensemble. Therefore, the present disclosure comprehends and reduces unnatural enhancements in human facial images, thereby having a significant impact on end users. With less memory and inference time, the present disclosure further allows real-time evaluation of AI based enhanced images.

[0027] In the following detailed description of the embodiments of the disclosure, reference is made to the accompanying drawings that form a part hereof, and in which are shown by way of illustration specific embodiments in which the disclosure may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the disclosure, and it is to be understood that other embodiments may be utilized and that changes may be made without departing from the scope of the present disclosure. The following description is, therefore, not to be taken in a limiting sense.

[0028] Figure 1 shows an embodiment of the present disclosure for determining naturalness of Artificial Intelligence (AI) based images.

[0029] Figure 1 shows an environment 100 for determining the naturalness of the AI based images. As shown, the environment 100 includes a naturalness determination system 101 connected through a communication network 105 to one or more sources 103. The one or more sources 103 may include, but are not limited to, image capturing sensors such as AI cameras, a database comprising AI based facial images of humans, and the like. The naturalness determination system 101 may include an Input / Output (I / O) interface 107, a memory 109 and a processor 111, as shown in Figure 1. The I / O interface 107 may receive an AI based facial image of a human, from the one or more sources 103. In an embodiment, the naturalness determination system 101 may be associated with a naturalness determination application which may be implemented on a personal device of a user. The naturalness determination system 101 may allow users to determine whether the AI based facial image of the user is natural or un-natural.

[0030] In an embodiment, the AI based facial image may be generated by using existing applications related to generation of AI images. In another embodiment, the AI based facial image may be captured by the AI cameras. The AI cameras are configured to enhance originally captured images by applying AI techniques to produce AI based images. The AI based facial image may be given as an input to the naturalness determination system 101 when the user may wish to verify the naturalness of the AI based images. The AI based facial image comprises one or more facial features which can be extracted from the AI based facial image. In an embodiment, the one or more facial features may include, but are not limited to, a mouth feature, a nose feature and an eye feature. The one or more facial features may be extracted from the AI based facial image using a pretrained AI model to identify specific facial artifacts. In an embodiment, the specific facial artifacts are identified based on facial landmarks and facial shape, identified from the AI based facial image. Therefore, based on the facial landmarks and the facial shape, patches respectively comprising the mouth feature, and the nose feature may be extracted. Furthermore, for the eye feature, the patches respectively comprising a left eye and a right eye are extracted and are further combined to generate a final patch comprising the eye feature. Thereafter, the one or more facial features as extracted are further evaluated for generation of confidence scores.

[0031] The naturalness determination system 101 is configured to obtain a feature confidence score for each of the one or more facial features and a facial confidence score for the AI based facial image. The naturalness determination system 101 is further configured to generate a feature confidence score for each of the one or more facial features and a facial confidence score for the AI based facial image. The feature confidence scores and the facial confidence score are generated simultaneously by using naturalness evaluation models. In an embodiment, feature confidence scores and the facial confidence score may be collectively referred as first confidence scores. In Machine Learning (ML), a confidence score corresponds to a number between 0 and 1 that represents a likelihood that an output generated by an AI or ML model is correct and will satisfy a user’s request. According to the present disclosure, feature-specific naturalness evaluation models are used for generating the respective feature confidence scores. The feature-specific naturalness evaluation models are trained for evaluating the respective facial features. For example, a first evaluation model may be trained for evaluating the mouth feature, a second evaluation model may be trained for evaluating the nose feature and a third evaluation model may be trained for evaluating the eye feature. Further, a face-specific naturalness evaluation model is used for generating the facial confidence score. The face-specific naturalness evaluation model is trained for evaluating a human face in the AI based facial image received by the naturalness determination system 101. For example, a fourth evaluation model may be trained for evaluating the human face in the AI based facial image. In an embodiment, the naturalness evaluation models are trained using a plurality of training AI based facial images. Herein, during the training, one or more facial features are extracted from the plurality of training AI based facial images. Further, the naturalness evaluation models are trained based on the naturalness associated with the plurality of training AI based facial images. In this manner, utilizing separate AI based naturalness evaluation models for each facial feature independently facilitates capturing minute artefacts on the human face.

[0032] Thereafter, the naturalness determination system 101 is configured to obtain a naturalness confidence score for the AI based facial image. In an embodiment, the naturalness determination system 101 is configured to determine a naturalness confidence score for the AI based facial image based on the feature confidence score for each of the one or more facial features and the facial confidence score. The naturalness confidence score for the AI based facial image may corresponds to a second confidence score. The second confidence score is determined based on a correlation of the feature confidence scores for each of the one or more facial features and the facial confidence score. Particularly, the second confidence score is determined based on the correlation of the first confidence scores. In an embodiment, the correlation of the first confidence scores is determined using at least one correlation model. In an embodiment, three correlation models corresponding to a first correlation model, a second correlation model and a third correlation model, may be used. Each of the three correlation models is trained based on predefined correlations associated with the human face. In an embodiment, the predefined correlations comprise correlations of the one or more facial features with various contexts including, but not limited to, age, gender, and a presence of an object on the human face, and the like. Therefore, the correlation models are independently trained for each of the predefined correlations associated with the human face. In an embodiment, the correlation models are independently trained based on the plurality of training AI based facial images. Particularly, the correlation models are trained based on a plurality of combinations of, the feature confidence scores and facial confidence scores, associated with the plurality of training AI based facial images. The correlation models are trained with respect to the predefined correlations associated with the human face. In an embodiment, the correlation models are configured to combine the first confidence score, with respect to the predefined correlations to determine the second confidence score. In an embodiment, the feature-specific naturalness evaluation models, the face-specific naturalness evaluation model and correlation models may be trained based on Convolutional Neural Network (CNN) architecture. In this manner, the present disclosure provides an efficient approach to amalgamate all confidence scores to derive a comprehensive confidence score to enhance overall model performance.

[0033] The naturalness determination system 101 is further configured to determine the naturalness of the AI based facial image by classifying the AI based facial image. The AI based facial images are classified as one of, natural image or un-natural image, based on the second confidence score. In an embodiment, the naturalness of the AI based facial image is determined by identifying whether the naturalness confidence score, corresponding to the second confidence score, is within a predefined threshold. In an embodiment, the predefined threshold is defined based on experimental analysis. For example, the predefined threshold may be set to a confidence score corresponding to a value near to 1. For instance, the predefined value may be set between 0.7-1.0. Therefore, when the second confidence score corresponds to a value below 0.7, then the AI based facial image is classified as unnatural. Whereas, when the second confidence score corresponds to a value between 0.7-1, then the AI based facial image is classified as natural. Therefore, the present disclosure proposes evaluation method with less memory and inference time, which allows addressing use-cases for performing real-time evaluation on enhanced facial images of the humans.

[0034] Figure 2 shows a detailed block diagram of a naturalness determination system in accordance with an embodiment of the present disclosure.

[0035] The naturalness determination system 101 may comprise memory 109 storing data 200 and one or more modules 202 which are described herein in detail. In an embodiment, data 200 may be stored within the memory 109. The data 200 may include, for example, input data 204, confidence score data 206, classification data 208 and other data 210. The data 200 may also include naturalness evaluation models (shown in Figure 3A) and correlation models (shown in Figure 3B).

[0036] The input data 204 may include data received from the one or more sources 103 that may comprise the AI based facial image of the human. The AI based facial image may correspond to enhanced images produced by the AI cameras.

[0037] The confidence score data 206 may include the feature confidence scores, the facial confidence score and the naturalness confidence score. In an embodiment, the feature confidence scores may include, but are not limited to, the first feature confidence score, the second feature confidence score and the third feature confidence score. The first feature confidence score may be determined for the first feature corresponding to the mouth feature. The second feature confidence score may be determined for the second feature corresponding to the nose feature. The third feature confidence score may be determined for the third feature corresponding to the eye feature. In an embodiment, the feature confidence scores and the facial confidence score may be referred as the first confidence score. In an embodiment, the naturalness confidence score may be referred to as the second confidence score.

[0038] The classification data 208 may include data related to the naturalness of the AI based facial image. In an embodiment, the naturalness may be classified by indicating whether the AI based facial image corresponds to a natural image or an un-natural image.

[0039] The other data 210 may store data, including temporary data and temporary files, generated by modules 202 for performing the various functions of the naturalness determination system 101.

[0040] In an embodiment, the data 200 in the memory 109 are processed by the one or more modules 202 present within the memory 109 of the naturalness determination system 101. In an embodiment, the one or more modules 202 may be implemented as dedicated units. As used herein, the term module refers to an application specific integrated circuit (ASIC), an electronic circuit, a field-programmable gate arrays (FPGA), Programmable System-on-Chip (PSoC), a combinational logic circuit, and / or other suitable components that provide the described functionality. In some implementations, the one or more modules 202 may be communicatively coupled to the processor 111 for performing one or more functions of the naturalness determination system 101. The said modules 202 when configured with the functionality defined in the present disclosure will result in a novel hardware.

[0041] In one implementation, the one or more modules 202 may include, but are not limited to, a receiving module 212, a score generation module 214, and a naturalness determination module 216. The one or more modules 202 may also include other modules 218 to perform various miscellaneous functionalities of the naturalness determination system 101.

[0042] The receiving module 212 may receive the input data 204 from the one or more sources 103. The user may feed an image in the receiving module 212 which corresponds to the AI based facial image. In an embodiment, the AI based facial image may be generated by using the existing applications related to generation of AI images. In an embodiment, the AI based facial image may be captured by the AI cameras. The AI cameras are configured to enhance the originally captured images by applying AI techniques to produce AI based images. In an embodiment, the user may capture the facial images of the humans using the AI cameras. The AI based facial image comprises the one or more facial features. In an embodiment, the one or more facial features may include, but are not limited to, the mouth feature, the nose feature, the eye feature, and the like. The receiving module 212 is further configured to extract the one or more facial features from the AI based facial image using the pretrained AI model to identify specific facial artifacts. The pretrained AI model may be a Convolutional Neural Network (CNN) model. In an embodiment, the pretrained AI model may be configured in the receiving module 212 for extracting the one or more facial features. The facial artifacts may include, but are not limited to, teeth in the mouth, colour of the eyes, length of hair, and the like. The pretrained AI model is trained to extract the one or more facial features by identifying the specific facial artifacts in predefined regions on the human face. The predefined regions may correspond to eye region, cheek region, forehead region, mouth region, and the like on the human face. Therefore, based on the facial artifacts identified in the predefined regions, patches respectively comprising the mouth feature and the nose feature are extracted by the receiving module 212. Furthermore, the receiving module 212 extracts the patches respectively comprising the left eye and the right eye for obtaining the eye feature and further combines the patches to generate the final patch comprising the eye feature. Thereafter, the receiving module 212 is configured to send the one or more features extracted from the input data 204 to the score generation module 214.

[0043] The score generation module 214 is configured to receive the extracted one or more features from the receiving module 212. The score generation module 214 is configured to generate the feature confidence score for each of the one or more facial features and the facial confidence score for the AI based facial image. In an embodiment, the feature confidence score and the facial confidence score may be collectively referred as first the confidence scores. The score generation module 214 generates the feature confidence scores by using feature-specific naturalness evaluation models. In an embodiment, the feature-specific naturalness evaluation models may include, but are not limited to, a mouth evaluation model, a nose evaluation model, an eyes evaluation model, and the like. The naturalness evaluation models may correspond to CNN models. The naturalness evaluation models are trained based on the one or more facial features that are extracted from the plurality of training AI based facial images, along with the naturalness associated with the plurality of training AI based facial images. For example, the mouth evaluation model may be trained based on a mouth feature extracted from each training AI based facial image from the plurality of training AI based facial images. Further, based on the mouth feature associated with each of the training AI based facial image, the corresponding training AI based facial images are labelled as one of, natural or un-natural. Thus, the mouth evaluation model is trained with the naturalness associated with each of the plurality of training AI based images.

[0044] The score generation module 214 generates the facial confidence score by using a face-specific naturalness evaluation model. According to the present disclosure, the feature-specific naturalness evaluation models are independently trained for specifically evaluating the respective facial features. Further, the face-specific naturalness evaluation model is trained for evaluating the human face in the AI based facial image as a single feature. The score generation module 214 relies on independent naturalness evaluation models for the generation of the first confidence scores for capturing minute artifacts on the human face. The score generation module 214 evaluates each feature of the one or more facial features on the human face in an effective manner by identifying natural and unnatural artefacts associated with the respective features. For example, the score generation module 214 may generate low feature confidence scores when the unnatural artefacts are identified in the one or more facial features of the human face. In an embodiment, the unnatural artefacts may correspond to, but are not limited to, improper construction of teeth, improper construction of eyes, mismatch in colour of the eyes, and the like. In this manner, individual confidence scores corresponding to the first confidence scores are generated by each of the naturalness evaluation models.

[0045] Thereafter, the score generation module 214 sends the first confidence scores to correlation models to determine the naturalness confidence score for the AI based facial image. The score generation module 214 determines the second confidence score based on the correlation of the first confidence scores and the predefined correlations associated with the human face. In an embodiment, the predefined correlations may comprise correlations of the one or more facial features with various contexts including but not limited to, age, gender, and a presence of an object on the human face, and the like. In an embodiment, the score generation module 214 may utilize individual correlation models for evaluating each correlation. The correlation models may correspond to CNN models.

[0046] For example, a first predefined correlation may correspond to correlating an adult laughing face with a baby’s laughing face. In another example, a second predefined correlation may correspond to correlating a male human face and a female human face with respect to facial hair. In a further example, a third correlation may correspond to correlating human faces with spectacles and the human faces without the spectacles. Therefore, based on the first predefined correlation, the score generation module 214 may first evaluate the human face in the AI based facial image corresponds to an adult face or a baby face. The score generation module 214 evaluates the human face as one of the baby face or the adult face, based on the evaluation of the AI based facial image and the evaluation of mouth feature of the human face. For example, consider that the human face may be evaluated as the baby face based on the evaluation of the AI based facial image, wherein an absence of hair and the facial hair may be observed. Further, the human face may be evaluated as the baby face based on the evaluation of the mouth feature, wherein an absence of teeth in the mouth feature may be observed. Thereafter, the score generation module 214 may determine the second confidence score 312 based on the correlation of the adult laughing face and the baby’s laughing face. For example, the score generation module 214 may determine a high naturalness confidence score upon evaluating the absence of the hair, the facial hair and the teeth in the AI based facial image. In another example, the score generation module 214 may determine a low naturalness confidence score upon evaluating the absence of the hair and the facial hair but presence of the teeth in the AI based facial image.

[0047] Thus, based on the predefined correlations with respect to the various contexts, the score generation module 214 efficiently determines the second confidence score. Thereafter, the score generation module 214 is configured to send the naturalness confidence score referred as the second confidence score to the naturalness determination module 216.

[0048] The naturalness determination module 216 is configured to receive the naturalness confidence score from the score generation module 214. The naturalness determination module 216 is configured to determine the naturalness of the AI based facial image. The naturalness determination module 216 is configured to classify the AI based facial image as the natural AI based facial image or the un-natural AI based facial image. The naturalness determination module 216 classifies the AI based facial image based on the naturalness confidence score. In an embodiment, the AI based facial images classified under the categories of natural AI based facial image and un-natural AI based facial image correspond to the classification data 208. Particularly, the naturalness determination module 216 is configured to identify whether the naturalness confidence score is within the predefined threshold. The predefined threshold is defined prior to performing the determination of the naturalness of the AI based facial image, based on experimental analysis. In an embodiment, the predefined threshold may be set to a confidence score corresponding to 0.8. According to this example, if the second confidence score determined by the naturalness determination module 216 corresponds to 0.7, then the naturalness of the AI based facial image is classified as non-natural. Subsequently, if the second confidence score determined by the naturalness determination module 216 corresponds to 0.9, then the naturalness of the AI based facial image is classified as natural.

[0049] Figure 3A and 3B show exemplary illustrations of determining naturalness of the AI based images using the naturalness evaluation models and the correlation models. As seen from Figure 3A, a facial image 302 may be fed into a feature extractor 304. The facial image 302 may correspond to the AI based facial image and the feature extractor 304 may be configured in the receiving module 212. The feature extractor 304 may extract the one or more facial features 306 from the facial image 302. As seen from Figure 3A, the one or more facial features 306 may comprise a first feature 3061 which may correspond to the mouth feature, a second feature 3062 which may correspond to the nose feature, and a third feature which may correspond to the eye feature. Thereafter, the one or more facial features 306 along with the facial image 302, are fed into the respective naturalness evaluation models 308. As seen from Figure 3A, the naturalness evaluation models 308 may include a first evaluation model 3081 for evaluating the mouth feature, a second evaluation model 3082 for evaluating the nose feature, a third evaluation model 3083 for evaluating the eye feature and a fourth evaluation model 3084 for evaluating the facial image 302. Thereafter, each of the naturalness evaluation models 308 is configured to generate the first confidence scores 310. Each of the naturalness evaluation models 308 is trained separately with its respective features for naturalness. For example, as seen from the Figure 3A, the first evaluation model 3081 may generate a first feature confidence score 3101 for the mouth feature. The second evaluation model 3082 may generate a second feature confidence score 3102 for the nose feature. The third evaluation model 3083 may generate a third feature confidence score 3103 for the eye feature. The fourth evaluation model 3084 may generate a fourth feature confidence score 3104 for the facial image 302. In an embodiment, the fourth evaluation model 3084 may consider the facial image 302 as a single feature for performing the evaluation. Subsequently, as seen from Figure 3B, the first confidence scores 310 are fed into the correlation models 312. The correlation models 312 may include a first correlation model 3121, a second correlation model 3122, and a third correlation model 3123. The correlation models 312 may take the first confidence scores 310 from the feature evaluator module as input and provides a second confidence score 314 as output. The correlation models 312 may use a method to merge the feature confidence scores from all facial features, resulting in a heightened and dependable confidence score. The correlation models 312 may integrate correlation scores from different machine learning correlation models optimized for specific artefacts, resulting in a comprehensive confidence score that is significantly more reliable due to model ensemble. The correlation models 312 may generate the comprehensive confidence score by comprehensively considering the presence, size, shape, form, color, and relationships of facial features.

[0050] In an embodiment, the first correlation model 3121 may correspond to correlating the adult laughing face with the baby’s laughing face. The first correlation model 3121 may be a model trained how facial muscles move when adults and babies laugh, how the structural characteristics of their expressions differ, and how the way emotions are conveyed changes. The first correlation model 3121 may analyze the facial features, expression styles, and characteristics of emotional expression that appear in the laughing faces of adults and babies. Correlating the adult laughing face with the baby’s laughing face may mean comparing and analyzing the laughing facial expression data of these two age groups to identify their correlations, such as similarities, differences, or patterns of continuity. The first correlation model 3121 may correspond to correlating the facial features based on the laughing face. The second correlation model 3122 may correspond to correlating the male human face and the female human face with respect to the facial hair of the human. The second correlation model 3122 may be a model trained on the presence, thickness, density, and distribution of facial hair in humans according to each gender or age group. Correlating the male and female human faces with respect to facial hair may mean comparing and analyzing the characteristics and differences of male and female faces, focusing on the presence, thickness, density, and distribution of facial hair (such as beards or fine hairs). The third correlation model 3123 may correspond to correlating the human faces with the spectacles and without the spectacles. The third correlation model 3123 may be a model trained to learn how a person’s facial features change when wearing spectacles versus when not wearing spectacles. Correlating the human face with spectacles and without spectacles may mean comparing and analyzing how a person’s facial features change when wearing spectacles versus when not wearing them. Therefore, the naturalness confidence score is determined by the correlation models 312. As seen from Figure 3B, the naturalness confidence score is referred as second confidence score 314.

[0051] Further, based on the second confidence score 314, the naturalness determination module 216 classifies the facial image 302 as one of a natural AI based facial image or an un-natural AI based facial image. Figure 4A and 4B show exemplary AI based images classified as the natural AI based facial image and the un-natural AI based facial image. Figure 4A shows AI based facial images classified as the natural AI based facial images. For example, as seen from Figure 4A, the facial images (3021, 3022, 3023) are classified as the natural AI based facial images. In another example, as seen from Figure 4B, the facial images (3024, 3025, 3026) are classified as the un-natural AI based facial images. From the example, the facial image 3024 may be classified as the un-natural AI based facial image. This is because the correlation models 312 may fail to identify the spectacles on the human, on scanning the facial image 3024 to identify features such as, a spectacle shape and uniform edges around the eyes, nose and face region. However, the correlation models 312 may succeed in identifying the features in the facial image 3021, and the facial image 3021 may therefore be classified as the natural AI based image. Similarly, the facial images 3025 and 3026 may be classified as the un-natural AI based images since the eye feature and the mouth feature of the human face are constructed in an improper manner. Particularly, in the facial image 3025, the eyes of the women are not properly constructed and therefore not visible, further the baby face in the facial image 3026 comprises teeth which is not natural. Whereas the facial image 3022 and 3023 may be classified as the natural AI based image since each facial feature is properly constructed. Particularly, in the facial image 3023, the mouth feature of the baby face does not comprise of the teeth, which is natural.

[0052] Figure 5A, 5B, and 5C show exemplary determination of naturalness using the correlation models 312. As seen from Figure 5A, when the natural image 3022 is fed into naturalness evaluation models 308, the first confidence scores generated by each evaluation model are between 0.7-0.9. For example, the first evaluation model 3081 may generate the first feature confidence score 3101 as 0.75, the second evaluation model 3082 may generate the second feature confidence score 3102 as 0.9, the third evaluation model 3083 may generate the third feature confidence score 3103 as 0.9, and the fourth evaluation model 3084 may generate the fourth feature confidence score 3104 as 0.85. Thereafter, based on the correlation of the first confidence scores 310 in the above example, the correlation models 312 may determine the second confidence score 314 which corresponds to 0.8. In an embodiment, if the predefined threshold for the naturalness confidence score is set to 0.7, then the facial image 3022 is classified as the natural AI based facial image of the human. Similarly, in Figure 5B and Figure 5C, the facial images (3024 and 3026) are classified as the un-natural facial images since the second confidence score 314 correspond to 0.2 and 0.1, respectively, which is below the predefined threshold. As observed from Figure 5B, though the first confidence scores 310 are higher (i.e., closer to the value 1), the second confidence score 314 is low. This is due to the correlation models 312 performing the correlation of the first confidence scores 310, with respect to the predefined correlations. For example, for the facial image 3024, the third correlation model 3123 may be used, which may correspond to correlating the human faces with the spectacles and without the spectacles. Since the construction of the spectacle on the human face in the facial image 3024 is not performed accurately, the second confidence score 314 is given as a low score corresponding to 0.2.

[0053] Figure 6 illustrates a flowchart showing a method for determining naturalness of Artificial Intelligence (AI) based images in accordance with an embodiment of present disclosure.

[0054] As illustrated in Figure 6, the method 600 comprises one or more blocks illustrating a method of determining naturalness of Artificial Intelligence (AI) generated images. The method 600 may be described in the general context of computer executable instructions. Generally, computer executable instructions can include routines, programs, objects, components, data structures, procedures, modules, and functions, which perform particular functions or implement particular abstract data types.

[0055] The order in which the method 600 is described is not intended to be construed as a limitation, and any number of the described method blocks can be combined in any order to implement the method. Additionally, individual blocks may be deleted from the methods without departing from the scope of the subject matter described herein. Furthermore, the method can be implemented in any suitable hardware, software, firmware, or combination thereof.

[0056] At block 601, the method 600 may include receiving, by the naturalness determination system 101, the AI based facial image of the human. The AI based facial image comprises the one or more facial features. The one or more facial features are extracted from the AI based facial image. The one or more facial features 306 of the human face may correspond to the mouth feature, the nose feature and the eye feature.

[0057] At block 603, the method 600 may include obtaining, by the naturalness determination system 101, the feature confidence score and the facial confidence score. The method 600 may include generating, by the naturalness determination system 101, the feature confidence score (3101, 3102, 3103) for each of the one or more facial features 306 using the feature specific naturalness evaluation models 308 which are trained for evaluating the respective facial features. The method 600 may further include generating the facial confidence score 3104 for the AI based facial image using the face-specific naturalness evaluation model trained for evaluating the human face. The feature confidence scores (3101, 3102, 3103) and the facial confidence score 3104 may be simultaneously generated. In an embodiment, the feature-specific naturalness evaluation models (3081, 3082, 3083) and the face-specific naturalness evaluation model 3084 are trained based on the naturalness associated with, the plurality of training AI based facial images and the one or more facial features 306 extracted from the plurality of training AI based facial images, respectively.

[0058] At block 605, the method 600 may include obtaining, by the naturalness determination system 101, the naturalness confidence score for the AI based facial image. The method 600 may include determining, by the naturalness determination system 101, the naturalness confidence score for the AI based facial image based on the correlation of the feature confidence score (3101, 3102, 3103) of each of the one or more facial features 306 and the facial confidence score 3104. In an embodiment, the correlation of the feature confidence score (3101, 3102, 3103) of each of the one or more facial features 306 and the facial confidence score 3104 are determined using the correlation models 312. The correlation models 312 are trained based on the plurality of combinations of, feature confidence scores and facial confidence scores associated with the plurality of training AI based facial images, with respect to the predefined correlations associated with the human face.

[0059] At block 607, the method 600 may include determining, by the naturalness determination system 101, the naturalness of the AI based facial image by classifying the AI based facial image based on the naturalness confidence score. In an embodiment, the method 600 may include determining, by the naturalness determination system 101,the naturalness of the AI based facial image by classifying the AI based facial image as one of, the natural AI based facial image or the un-natural AI based facial image, based on the naturalness confidence score referred as the second confidence score 314. In an embodiment, identifying the naturalness of the AI based facial image comprises identifying whether the naturalness confidence score is within the predefined threshold.

[0060] Figure 7 is a block diagram of an exemplary computer system for implementing an embodiment of the present disclosure.

[0061] In some embodiments, Figure 7 illustrates a block diagram of an exemplary computer system 700 for implementing embodiments consistent with the present disclosure. In some embodiments, the computer system 700 can be the naturalness determination system 101 that comprises a processor (also referred as a processor 702 in this Figure 7) that is used for determining the naturalness of the AI based facial image. The processor 702 may include at least one data processor for executing program components for executing user or system-generated business processes. The processor 702 may include specialized processing units such as integrated system (bus) controllers, memory management control units, floating point units, graphics processing units, digital signal processing units, etc.

[0062] The processor 702 may be disposed in communication with input devices 710 and output devices 711 via I / O interface 701. The I / O interface 701 may employ communication protocols / methods such as, without limitation, audio, analog, digital, stereo, IEEE-1394, serial bus, Universal Serial Bus (USB), infrared, PS / 2, BNC, coaxial, component, composite, Digital Visual Interface (DVI), high-definition multimedia interface (HDMI), Radio Frequency (RF) antennas, S-Video, Video Graphics Array (VGA), IEEE 802.n / b / g / n / x, Bluetooth, cellular (e.g., Code-Division Multiple Access (CDMA), High-Speed Packet Access (HSPA+), Global System For Mobile Communications (GSM), Long-Term Evolution (LTE), WiMax, or the like), microphones etc.

[0063] Using the I / O interface 701, computer system 700 may communicate with input devices 710 and output devices 711.

[0064] In some embodiments, the processor 702 may be disposed in communication with a communication network 709 via a network interface 703. The network interface 703 may communicate with the communication network 709. The network interface 703 may employ connection protocols including, without limitation, direct connect, Ethernet (e.g., twisted pair 10 / 100 / 1000 Base T), Transmission Control Protocol / Internet Protocol (TCP / IP), token ring, IEEE 802.11a / b / g / n / x, OBD-II (On-Board Diagnostics) connector, etc. Using the network interface 703 and the communication network 709, the computer system 700 may communicate with one or more sources 103.

[0065] The communication network 709 can be implemented as one of the different types of networks, such as intranet or Local Area Network (LAN) and such within the organization. The communication network 709 may either be a dedicated network or a shared network, which represents an association of the different types of networks that use a variety of protocols, for example, Hypertext Transfer Protocol (HTTP), Transmission Control Protocol / Internet Protocol (TCP / IP), Wireless Application Protocol (WAP), etc., to communicate with each other.

[0066] Further, the communication network 709 may include a variety of network devices, including routers, bridges, servers, computing devices, storage devices, etc. In some embodiments, the processor 702 may be disposed in communication with a memory 705 (e.g., RAM, ROM, etc. not shown in Figure 7) via a storage interface 704. The storage interface 704 may connect to memory 705 including, without limitation, memory drives, removable disc drives, etc., employing connection protocols such as Serial Advanced Technology Attachment (SATA), Integrated Drive Electronics (IDE), IEEE-1394, Universal Serial Bus (USB), fibre channel, Small Computer Systems Interface (SCSI), etc. The memory drives may further include a drum, magnetic disc drive, magneto-optical drive, optical drive, Redundant Array of Independent Discs (RAID), solid-state memory devices, solid-state drives, etc.

[0067] The memory 705 may store a collection of program or database components, including, without limitation, a user interface 706, an operating system 707, a web browser 708 etc. In some embodiments, the computer system 700 may store user / application data, such as the data, variables, records, etc. as described in this present disclosure. Such databases may be implemented as fault-tolerant, relational, scalable, secure databases such as Oracle or Sybase.

[0068] Operating system 707 may facilitate resource management and operation of computer system 700. Examples of operating systems include, without limitation, APPLE® MACINTOSH® OS X®, UNIX®, UNIX-like system distributions (E.G., BERKELEY SOFTWARE DISTRIBUTION® (BSD), FREEBSD®, NETBSD®, OPENBSD, etc.), LINUX® DISTRIBUTIONS (E.G., RED HAT®, UBUNTU®, KUBUNTU®, etc.), IBM®OS / 2®, MICROSOFT® WINDOWS® (XP®, VISTA® / 7 / 8, 10 etc.), APPLE® IOS®, GOOGLETM ANDROIDTM, BLACKBERRY® OS, or the like. User interface 706 may facilitate display, execution, interaction, manipulation, or operation of program components through textual or graphical facilities. For example, user interfaces may provide computer interaction interface elements on a display system operatively connected to computer system 700, such as cursors, icons, check boxes, menus, scrollers, windows, widgets, etc. Graphical User Interfaces (GUIs) may be employed, including, without limitation, Apple® Macintosh® operating systems’ Aqua®, IBM® OS / 2®, Microsoft® Windows® (e.g., Aero, Metro, etc.), web interface libraries (e.g., ActiveX®, Java®, Javascript®, AJAX, HTML, Adobe® Flash®, etc.), or the like.

[0069] Computer system 700 may implement web browser 708 stored program components. Web browser 708 may be a hypertext viewing application, such as MICROSOFT® INTERNET EXPLORER®, GOOGLETM CHROMETM, MOZILLA® FIREFOX®, APPLE® SAFARI®, etc. Secure web browsing may be provided using Secure Hypertext Transport Protocol (HTTPS), Secure Sockets Layer (SSL), Transport Layer Security (TLS), etc. Web browsers 708 may utilize facilities such as AJAX, DHTML, ADOBE® FLASH®, JAVASCRIPT®, JAVA®, Application Programming Interfaces (APIs), etc. Computer system 700 may implement a mail server stored program component. The mail server may be an Internet mail server such as Microsoft Exchange, or the like. The mail server may utilize facilities such as ASP, ACTIVEX®, ANSI® C++ / C#, MICROSOFT®, NET, CGI SCRIPTS, JAVA®, JAVASCRIPT®, PERL®, PHP, PYTHON®, WEBOBJECTS®, etc. The mail server may utilize communication protocols such as Internet Message Access Protocol (IMAP), Messaging Application Programming Interface (MAPI), MICROSOFT® exchange, Post Office Protocol (POP), Simple Mail Transfer Protocol (SMTP), or the like. In some embodiments, the computer system 700 may implement a mail client stored program component. The mail client may be a mail viewing application, such as APPLE® MAIL, MICROSOFT® ENTOURAGE®, MICROSOFT® OUTLOOK®, MOZILLA® THUNDERBIRD®, etc.

[0070] Furthermore, one or more computer-readable storage media may be utilized in implementing embodiments consistent with the present disclosure. A computer-readable storage medium refers to any type of physical memory on which information or data readable by a processor may be stored. Thus, a computer-readable storage medium may store instructions for execution by one or more processors, including instructions for causing the processor(s) to perform steps or stages consistent with the embodiments described herein. The term “computer-readable medium” should be understood to include tangible items and exclude carrier waves and transient signals, i.e., non-transitory. Examples include Random Access Memory (RAM), Read-Only Memory (ROM), volatile memory, non-volatile memory, hard drives, Compact Disc (CD) ROMs, Digital Video Disc (DVDs), flash drives, disks, and any other known physical storage media.

[0071] An embodiment of the present disclosure facilitates determination of the naturalness of the AI based images based on confidence scores generated for each feature in the AI based image. Particularly, the confidence scores are generated by individual feature specific models, that are trained to intelligently merge confidence score from various facial features into a single comprehensive confidence score. The resulting comprehensive confidence score after amalgamating all confidence score is significantly more reliable due to model ensemble. Therefore, the present disclosure comprehends and reduces unnatural enhancements in human facial images, thereby having a significant impact on end users. With less memory and inference time, the present disclosure further allows real-time evaluation of AI based enhanced images.

[0072] According to an embodiment of the disclosure, a method of determining naturalness of Artificial Intelligence (AI) based images is provided. The method may include receiving, by a naturalness determination system 101, an AI based facial image of a human. The AI based facial image may comprise one or more facial features. The one or more facial features may be extracted from the AI based facial image. The method may include obtaining, by the naturalness determination system 101, a feature confidence score for each of the one or more facial features using feature-specific naturalness evaluation models trained for evaluating the respective facial features, and a facial confidence score for the AI based facial image using a face-specific naturalness evaluation model trained for evaluating a human face. The method may include obtaining, by the naturalness determination system 101, a naturalness confidence score for the AI based facial image based on a correlation of the feature confidence score for each of the one or more facial features and the facial confidence score. The method may include determining, by the naturalness determination system 101, naturalness of the AI based facial image by classifying the AI based facial image based on the naturalness confidence score.

[0073] According to an embodiment of the disclosure, the one or more facial features of the human face may correspond to a mouth feature, a nose feature, and an eye feature.

[0074] According to an embodiment of the disclosure, the feature-specific naturalness evaluation models and the face-specific naturalness evaluation model may be trained based on the naturalness associated with a plurality of training AI based facial images and the one or more facial features extracted from the plurality of training AI based facial images, respectively.

[0075] According to an embodiment of the disclosure, the correlation between the feature confidence score for each of the one or more facial features and the facial confidence score may be determined using correlation models.

[0076] According to an embodiment of the disclosure, the correlation models may be trained based on a plurality of combinations of feature confidence scores and facial confidence scores associated with a plurality of training AI based facial images, with respect to predefined correlations associated with the human face.

[0077] According to an embodiment of the disclosure, the correlation models may be trained for each of the predefined correlations associated with the human face.

[0078] According to an embodiment of the disclosure, the predefined correlations may relate to facial features with respect to age, gender, a presence of an object on the human face.

[0079] According to an embodiment of the disclosure, the feature-specific naturalness evaluation models, the face-specific naturalness evaluation model, and correlation models may be trained based on Convolutional Neural Network (CNN) architecture.

[0080] According to an embodiment of the disclosure, the determining of the naturalness of the AI based facial image may comprise identifying whether the naturalness confidence score is within a predefined threshold.

[0081] According to an embodiment of the disclosure, a naturalness determination system 101 for determining naturalness of Artificial Intelligence (AI) generated images is provided. The naturalness determination system 101 may include a processor 111 and a memory 109, storing instructions, communicatively coupled to the processor 111. The instructions, when executed by the processor 111, may cause the system 100 to receive an AI based facial image of a human. The AI based facial image may comprise one or more facial features. The one or more facial features may be extracted from the AI based facial image. The instructions, when executed by the processor 111, may cause the system 100 to obtain a feature confidence score for each of the one or more facial features using feature-specific naturalness evaluation models trained for evaluating the respective facial features, and a facial confidence score for the AI based facial image using a face-specific naturalness evaluation model trained for evaluating a human face. The instructions, when executed by the processor 111, may cause the system 100 to obtain a naturalness confidence score for the AI based facial image based on a correlation of the feature confidence score for each of the one or more facial features and the facial confidence score. The instructions, when executed by the processor 111, may cause the system 100 to determine naturalness of the AI based facial image by classifying the AI based facial image based on the naturalness confidence score.

[0082] According to an embodiment of the disclosure, the one or more facial features of the human face may correspond to a mouth feature, a nose feature, and an eye feature.

[0083] According to an embodiment of the disclosure, the processor 111 may be configured to train the feature-specific naturalness evaluation models and the face-specific naturalness evaluation model based on the naturalness associated with a plurality of training AI based facial images and the one or more facial features extracted from the plurality of training AI based facial images, respectively.

[0084] According to an embodiment of the disclosure, the processor 111 may be configured to determine the correlation between the feature confidence score for each of the one or more facial features and the facial confidence score using correlation models.

[0085] According to an embodiment of the disclosure, the processor 111 may be configured to train the correlation models based on a plurality of combinations of feature confidence scores and facial confidence scores associated with a plurality of training AI based facial images, with respect to predefined correlations associated with the human face.

[0086] According to an embodiment of the disclosure, the correlation models may be trained for each of the predefined correlations associated with the human face.

[0087] According to an embodiment of the disclosure, the predefined correlations may relate to facial features with respect to age, gender, a presence of an object on the human face.

[0088] According to an embodiment of the disclosure, the processor 111 may be configured to train the feature-specific naturalness evaluation models, the face-specific naturalness evaluation model, and correlation models based on Convolutional Neural Network (CNN) architecture.

[0089] According to an embodiment of the disclosure, the processor 111 may be configured to determine the naturalness of the AI based facial image by identifying whether the naturalness confidence score is within a predefined threshold.

[0090] According to an embodiment of the disclosure, a computer-readable storage medium storing instructions is provided. The instructions, when executed by the at least one processor, may cause the at least one processor to receive an AI based facial image of a human. The AI based facial image may comprise one or more facial features. The one or more facial features may be extracted from the AI based facial image. The instructions, when executed by the at least one processor, may cause the at least one processor to obtain a feature confidence score for each of the one or more facial features using feature-specific naturalness evaluation models trained for evaluating the respective facial features, and a facial confidence score for the AI based facial image using a face-specific naturalness evaluation model trained for evaluating a human face. The instructions, when executed by the at least one processor, may cause the at least one processor to obtain a naturalness confidence score for the AI based facial image based on a correlation of the feature confidence score for each of the one or more facial features and the facial confidence score. The instructions, when executed by the at least one processor, may cause the at least one processor to determine naturalness of the AI based facial image by classifying the AI based facial image based on the naturalness confidence score.

[0091] With respect to the use of substantially any plural and / or singular terms herein, those having skill in the art can translate from the plural to the singular and / or from the singular to the plural as is appropriate to the context and / or application. The various singular / plural permutations may be expressly set forth herein for sake of clarity.

[0092] It will be understood by those within the art that, in general, terms used herein, and especially in the appended claims (e.g., bodies of the appended claims) are generally intended as “open” terms (e.g., the term “including” should be interpreted as “including but not limited to,” the term “having” should be interpreted as “having at least,” the term “includes” should be interpreted as “includes but is not limited to,” etc.).

[0093] It will be further understood by those within the art that if a specific number of an introduced claim recitation is intended, such an intent will be explicitly recited in the claim, and in the absence of such recitation no such intent is present. For example, as an aid to understanding, the following appended claims may contain usage of the introductory phrases “at least one” and “one or more” to introduce claim recitations.

[0094] However, the use of such phrases should not be construed to imply that the introduction of a claim recitation by the indefinite articles “a” or “an” limits any particular claim containing such introduced claim recitation to present disclosure containing only one such recitation, even when the same claim includes the introductory phrases “one or more” or “at least one” and indefinite articles such as “a” or “an” (e.g., “a” and / or “an” should typically be interpreted to mean “at least one” or “one or more”); the same holds true for the use of definite articles used to introduce claim recitations. In addition, even if a specific number of an introduced claim recitation is explicitly recited, those skilled in the art will recognize that such recitation should typically be interpreted to mean at least the recited number (e.g., the bare recitation of “two recitations,” without other modifiers, typically means at least two recitations, or two or more recitations).

[0095] Furthermore, in those instances where a convention analogous to “at least one of A, B, and C, etc.” is used, in general such a construction is intended in the sense one having skill in the art would understand the convention (e.g., “a system having at least one of A, B, and C” would include but not be limited to systems that have A alone, B alone, C alone, A and B together, A and C together, B and C together, and / or A, B, and C together, etc.). In those instances where a convention analogous to “at least one of A, B, or C, etc.” is used, in general such a construction is intended in the sense one having skill in the art would understand the convention (e.g., “a system having at least one of A, B, or C” would include but not be limited to systems that have A alone, B alone, C alone, A and B together, A and C together, B and C together, and / or A, B, and C together, etc.). It will be further understood by those within the art that virtually any disjunctive word and / or phrase presenting two or more alternative terms, whether in the description, claims, or drawings, should be understood to contemplate the possibilities of including one of the terms, either of the terms, or both terms. For example, the phrase “A or B” will be understood to include the possibilities of “A” or “B” or “A and B”.

[0096] In addition, where features or aspects of the disclosure are described in terms of Markush groups, those skilled in the art will recognize that the disclosure is also thereby described in terms of any individual member or subgroup of members of the Markush group.

[0097] While various aspects and embodiments have been disclosed herein, other aspects and embodiments will be apparent to those skilled in the art. The various aspects and embodiments disclosed herein are for purposes of illustration and are not intended to be limiting, with the true scope and spirit being indicated by the following claims.

Claims

1.A method (600) of determining naturalness of Artificial Intelligence (AI) based images, the method comprising:receiving (601), by a naturalness determination system (101), an AI based facial image of a human, the AI based facial image comprising one or more facial features extracted from the AI based facial image;obtaining (603), by the naturalness determination system (101):a feature confidence score for each of the one or more facial features using feature-specific naturalness evaluation models trained for evaluating the respective facial features, anda facial confidence score for the AI based facial image using a face-specific naturalness evaluation model trained for evaluating a human face;obtaining (605), by the naturalness determination system (101), a naturalness confidence score for the AI based facial image based on a correlation of the feature confidence score for each of the one or more facial features and the facial confidence score; anddetermining (607), by the naturalness determination system (101), naturalness of the AI based facial image by classifying the AI based facial image based on the naturalness confidence score.2.The method as claimed in claim 1, wherein the one or more facial features of the human face correspond to a mouth feature, a nose feature, and an eye feature.3.The method as claimed in any one of claims 1 to 2, wherein the feature-specific naturalness evaluation models and the face-specific naturalness evaluation model are trained based on the naturalness associated with a plurality of training AI based facial images and the one or more facial features extracted from the plurality of training AI based facial images, respectively.4.The method as claimed in any one of claims 1 to 3, wherein the correlation between the feature confidence score for each of the one or more facial features and the facial confidence score is determined using correlation models.5.The method as claimed in any one of claims 1 to 4, wherein the correlation models are trained based on a plurality of combinations of, feature confidence scores and facial confidence scores associated with a plurality of training AI based facial images, with respect to predefined correlations associated with the human face.6.The method as claimed in any one of claims 1 to 5, wherein the correlation models are trained for each of the predefined correlations associated with the human face.7.The method as claimed in any one of claims 1 to 6, wherein the predefined correlations relate to facial features with respect to age, gender, a presence of an object on the human face.8.The method as claimed in any one of claims 1 to 7, wherein the feature-specific naturalness evaluation models, the face-specific naturalness evaluation model, and correlation models are trained based on Convolutional Neural Network (CNN) architecture.9.The method as claimed in any one of claims 1 to 8, wherein the determining of the naturalness of the AI based facial image comprises identifying whether the naturalness confidence score is within a predefined threshold.10.A naturalness determination system (101) for determining naturalness of Artificial Intelligence (AI) generated images, comprising:a processor (111); anda memory (109) communicatively coupled to the processor (111), wherein the memory (109) stores processor instructions, which, on execution, causes the processor (111) to:receive an AI based facial image of a human, generated using AI, the AI based facial image comprising one or more facial features extracted from the AI based facial image;obtain:a feature confidence score for each of the one or more facial features using feature-specific naturalness evaluation models trained for evaluating the respective facial features, anda facial confidence score for the AI based facial image using a face-specific naturalness evaluation model trained for evaluating a human face;obtain a naturalness confidence score for the AI based facial image based on a correlation of, the feature confidence score for each of the one or more facial features and the facial confidence score; anddetermine naturalness of the AI based facial image by classifying the AI based facial image based on the naturalness confidence score.11.The naturalness determination system (101) as claimed in claim 10, wherein the one or more facial features of the human face correspond to a mouth feature, a nose feature, and an eye feature.12.The naturalness determination system (101) as claimed in any one of claims 10 to 11, wherein the processor is configured to train the feature-specific naturalness evaluation models and the face-specific naturalness evaluation model based on the naturalness associated with, a plurality of training AI based facial images and the one or more facial features extracted from the plurality of training AI based facial images, respectively.13.The naturalness determination system (101) as claimed in any one of claims 10 to 12, wherein the processor is configured to determine the correlation between the feature confidence score for each of the one or more facial features and the facial confidence score using correlation models.14.The naturalness determination system (101) as claimed in any one of claims 10 to 13, wherein the processor is configured to train the correlation models based on a plurality of combinations of, feature confidence scores and facial confidence scores associated with a plurality of training AI based facial images, with respect to predefined correlations associated with the human face.15.A computer-readable storage medium storing instructions that, when executed by at least one processor, cause the at least one processor to:receive an AI based facial image of a human, generated using AI, the AI based facial image comprising one or more facial features extracted from the AI based facial image;obtain:a feature confidence score for each of the one or more facial features using feature-specific naturalness evaluation models trained for evaluating the respective facial features, anda facial confidence score for the AI based facial image using a face-specific naturalness evaluation model trained for evaluating a human face;obtain a naturalness confidence score for the AI based facial image based on a correlation of the feature confidence score for each of the one or more facial features and the facial confidence score; anddetermine naturalness of the AI based facial image by classifying the AI based facial image based on the naturalness confidence score.

Images