Determination of product authenticity

1 . A method of determining authenticity of a product, the method comprising:

(a) acquiring image data, using an image sensor, indicative of at least one image including a label of the product, the label including information related to the product;

(b) identifying a plurality of regions of the label that include information related to the product, and classifying each identified region based on a determined form of information included in said region, wherein the determined form of information includes text, barcodes, images, Quick Response (QR) codes or logos, the identification and classification steps being performed by executing a trained machine learning model based on the acquired image data, wherein classifying the identified regions of the label comprises determining coordinates defining a position of each identified region within the label, wherein at least two regions of the label have a different classification, wherein identifying the plurality of regions comprises:

normalizing a projection of the acquired image data, the normalizing comprising:

identifying coordinates of the label; and

scaling the identified coordinates such that a rectangular region of the label is obtained;

(c) for each classified region of the label, identifying one or more attributes of said classified region based on the acquired image data, and comparing the identified attributes against a selected set of one or more attributes from a plurality of expected attributes, wherein the plurality of expected attributes correspond to a label region of an authentic product that includes the determined form of information of said classified region,

wherein the selected set of one or more attributes is derived from the plurality of expected attributes based on the one or more attributes of said classified region,

wherein the one or more attributes comprise the position of the classified region within the label, and

wherein the comparison of the identified attributes comprises comparing the position of the classified region relative to an expected position of the respective label region of the authentic product that includes the determined form of information of the classified region; and

(d) providing an indication of authenticity of the product determined based on the comparison in the form of an audio-visual warning.

2 . The method according to claim 1 , wherein the machine learning model is trained using a training dataset comprising a plurality of training samples each including product label image data and associated classified regions indicating the type of information in the respective regions.

3 . The method according to claim 2 , wherein the training dataset includes training samples related to authentic product labels, training samples related to counterfeit product labels, or a combination of both.

4 . The method according to claim 1 , wherein the attributes include at least one among:

content of the information of the classified region within the label, and wherein the comparison step includes comparing the content of the classified region relative to expected content of a label region of an authentic product that includes the determined type of information of said classified region; and

one or more display characteristics of content in the classified region, and wherein the comparison step includes comparing the display characteristics of the classified region relative to expected display characteristics of a label region of an authentic product that includes the determined type of information of said classified region.

5 . The method according to claim 4 , wherein the content includes at least one of: a serial number; a date code; a catalogue number; a UL (Underwriters Laboratories) certification; product amperage; product temperature rise; product style number; number of poles of products; and a UPC (Universal Product Code) number.

6 . The method according to claim 1 , wherein the attributes include one or more display characteristics of content in the classified region, wherein the comparison step includes comparing the display characteristics of the classified region relative to expected display characteristics of a label region of an authentic product that includes the determined type of information of said classified region and wherein the display characteristics include at least one of colour, size, font, resolution, and fading of the content, and wherein the content includes one or more alphanumeric or special characters, or a logo or image.

7 . The method according to claim 1 , the method comprising:

identifying coordinates in the acquired image data that define edges of the label;

determining a scaling factor to be applied to the coordinates to obtain a rectangular area defining the label; and

determining an expected error in the step of identifying the attributes of each classified region,

wherein the step of identifying the attributes comprises determining a confidence score associated with the identification, and

wherein the indication of authenticity is determined based on the confidence score relative to the expected error.

8 . The method according to claim 1 , the method comprising:

scanning, using a scanning device, a barcode on the label; and,

accessing a database to check whether the barcode corresponds to an authentic product, wherein the indication of authenticity of the product is determined based on the barcode check.

9 . The method according to claim 1 , the method comprising identifying a category of the product, based on the acquired image data, by accessing a database storing image data relating to a plurality of potential product categories, wherein the expected attributes for each label region are based on the identified product category.

10 . The method according to claim 1 , wherein the acquired image data includes data indicative of two or more labels of the product, the method comprising performing steps (b), (c) and (d) for each of the two or more labels, and the method comprising providing an overall indication of authenticity of the product determined based on the respective indication of authenticity for each of the two or more labels.

11 . The method according to claim 1 , the method comprising:

acquiring sensor data indicative of an exterior of the product;

identifying a category of the product, from a plurality of potential product categories, by executing a trained machine learning model based on the acquired sensor data;

identifying one or more attributes of the product based on the acquired sensor data, and comparing the identified attributes against the selected set of one or more attributes for an authentic product in the identified category; and,

providing an indication of authenticity of the product determined based on the comparison.

12 . The method according to claim 11 , wherein the one or more attributes includes a colour of at least one part of the product, and wherein the comparison step includes comparing the colour determined from the sensor data against an expected threshold colour range of the at least one part for an authentic product, and wherein the at least one part is a logo or trademark included in the sensor data, or a product housing included in the sensor data.

13 . The method according to claim 11 , wherein the one or more attributes include one or more dimensions of the product, and wherein the comparison step includes comparing the dimensions determined from the sensor data against expected dimensions for an authentic product.

14 . The method according to claim 12 , wherein the sensor data includes visual data, acquired from a visual sensor, indicative of a visual of the exterior of the product, and wherein the visual sensor is the image sensor.

15 . The method according to claim 11 , wherein the sensor data includes radar data, acquired from a radar sensor, wherein the one or more attributes includes at least one among

a shape of the product, and wherein the comparison step includes comparing the shape determined from the sensor data against an expected shape for an authentic product; and

a roughness of an exterior surface of the product, and wherein the comparison step includes comparing the roughness determined from the sensor data against an expected surface roughness for an authentic product.

16 . The method according to claim 11 , the method comprising providing an overall indication of authenticity of the product determined based on the respective determined indications of authenticity.

17 . The method according to claim 1 , the method further comprising:

acquiring sensor data indicative of an exterior of the product;

identifying a category of the product, from a plurality of potential product categories, by executing a trained machine learning model based on the acquired sensor data;

identifying one or more attributes of the product based on the acquired sensor data, and comparing the identified attributes against the selected set of one or more attributes for an authentic product in the identified category; and,

providing an indication of authenticity of the product determined based at least in part on the comparison.

18 . The method according to claim 1 , wherein determining the indication of authenticity of the product comprises:

calculating a confidence score that the product is counterfeit, based on the comparison; and

providing the audio-visual warning signal indicating that the product is counterfeit, based on determining that the calculated confidence score is greater than a preterminal threshold.

19 . The method according to claim 1 , wherein identifying the one or more attributes of the classified region comprises:

determining a relative location of the classified region in the label, wherein determining the relative location of the classified region comprises:

determining a width to length ratio of the label;

determine coordinates of a set of central pixels of the label;

determining a centroid region of the classified region and calculating a distance of the centroid region from the set of central pixels; and

normalizing the distance based on the width to length ratio.

20 . A portable device for determining authenticity of a product, the portable device comprising:

an image sensor configured to acquire image data indicative of at least one image including a label of the product, the label including information related to the product;

a processor configured to:

identify a plurality of regions of the label that include information related to the product, and classify each identified region based on a determined form of information included in said region, wherein the determined form of information includes text, barcodes, images, Quick Response (QR) codes or logos, the identification and classification steps being performed by executing a trained machine learning model based on the acquired image data, wherein classifying the identified regions of the label comprises determining coordinates defining a position of each identified region within the label, wherein at least two regions of the label have a different classification, wherein identifying the plurality of regions comprises:

normalizing a projection of the acquired image data, the normalizing comprising:

identifying coordinates of the label; and

scaling the identified coordinates such that a rectangular region of the label is obtained; and

for each classified region of the label, identify one or more attributes of said classified region based on the acquired image data, and compare the identified attributes against a selected set of one or more attributes from a plurality of expected attributes, wherein the plurality of expected attributes correspond to a label region of an authentic product that includes the determined form of information of said classified region,

wherein the selected set of one or more attributes is derived from the plurality of expected attributes based on the one or more attributes of said classified region,

wherein the one or more attributes comprise the position of the classified region within the label, and

wherein the comparison of the identified attributes comprises comparing the position of the classified region relative to an expected position of the respective label region of the authentic product that includes the determined form of information of the classified region; and

an output configured to provide an indication of authenticity of the product determined based on the comparison in the form of an audio-visual warning.