System and method for image processing for trend analysis

1. A method for analyzing trends using image data, comprising:

generating visual tokens associated with images from a plurality of image data sets, wherein generating the visual tokens comprises, for each image in the plurality of image data sets:

detecting an object in the image by performing foreground detection on the image to determine contours of the object;

determining feature descriptors of the image;

classifying the object in accordance with an object category;

determining one or more attributes of the object, based on the feature descriptors within the contours of the object; and

generating a visual token associated with the image, the visual token comprising information including the object category and the one or more attributes of the object determined from the feature descriptors;

storing the visual tokens in a token database, wherein each visual token stored in the database is stored in association with an identifier that identifies the visual token as being associated with an image data set;

receiving a trend query relating to an object category and defining two or more groups of interest for which trend data is to be determined with a defined time period;

identifying visual tokens of interest relating to the object category for use in searching the token database storing the plurality of visual tokens;

identifying one or more image data sets associated with each of the two or more groups of interest;

analyzing visual tokens matching the visual tokens of interest that are associated with the identified image data sets to generate trend data for each of the two or more groups of interest for the defined period of time; and

generating an output based on a comparison of the trend data generated for the two or more groups of interest for the defined period of time.

2. The method of claim 1 , wherein the trend query comprises the visual tokens of interest, or at least one of an image and one or more search terms from which the visual tokens of interest can be derived.

3. The method of claim 1 , further comprising retrieving images from the one or more image data sets within the defined time period, and generating one or more visual tokens for each of the images.

4. The method of claim 1 , wherein the trend data generated by analyzing the visual tokens is indicative of a presence of an object attribute in relation to the object category over time.

5. The method of claim 4 , wherein the trend query further defines the object attribute for which trend data is to be determined, and wherein the visual tokens of interest relate to the object attribute.

6. The method of claim 1 , wherein each of the two or more groups of interest comprise one of: an influencer, a group of influencers, a retailer, a group of retailers, a brand, and a group of brands.

7. The method of claim 1 , wherein the plurality of image data sets are retrieved from one or more of: the Internet, social media sites, television content, and product databases.

8. The method of claim 1 , wherein classifying the object comprises:

computing, with a simple cells layer, convolution operations using filters selected from a group consisting of: 2D Gabor, or Gabor-like filters, over a range of parameters;

performing, with a complex cells layer, a pooling operation over a retinotopically local cluster of simple cells within a complex receptive field;

sending an output of the pooling operation to one of a group consisting of: the simple cells layer, another complex cells layer, or a feature descriptor layer;

terminating computations in the simple cells layer and complex cells layer as per a Cellular Neural Network definition, wherein the output is a set of values representing a set of feature descriptors;

assessing, with a complex localization cells layer, regions in which the object of the product image is located; and

classifying the object using the complex localization cells layer, regions and feature descriptors.

9. The method of claim 1 , wherein the token database further stores a plurality of textual tokens, wherein each textual token stored in the token database is stored in association with an identifier that identifies the textual token as being associated with an image data set.

10. The method of claim 9 , further comprising populating the database with the plurality of textual tokens, wherein the plurality of textual tokens are generated from a plurality of images, and for a respective image a textual token is generated by:

determining image metadata for the image; and

generating the textual token associated with the image, the textual token comprising the image metadata.

11. The method of claim 10 , wherein the image metadata comprises one or more of:

a timestamp of the image, descriptive details associated with the image, a popularity of the image, and a location of the image, and wherein the descriptive details comprise one or more of: a social network, a user name, a caption, a URL associated with the image, pixel co-ordinates associated with products or brands present in the image, and details about a subject in the image.

12. The method of claim 11 , wherein the trend query defines grouping parameters specifying the two or more groups of interest, and wherein identifying the one or more image data sets associated with each of the two or more groups of interest is performed by analyzing the image metadata of the textual tokens based on the grouping parameters.

13. The method of claim 11 , further comprising identifying textual tokens of interest relating to the object category in the trend query, and wherein generating the trend data for each of the two or more groups of interest is further based on an analysis of textual tokens matching the textual tokens of interest that are associated with each of the identified image data sets.

14. The method of claim 1 , further comprising displaying the output on a display.

15. The method of claim 1 , further comprising generating an activation for the trend, the activation comprising one or more scheduled activities relating to the trend data.

16. A system for analyzing trends using image data, comprising:

a token database storing a plurality of visual tokens, wherein each visual token stored in the token database is stored in association with an identifier that identifies the visual token as being associated with an image data set; and

a server comprising at least one processor configured to:

generate the plurality of visual tokens for images from a plurality of image data sets,

wherein generating the plurality of visual tokens comprises, for each image in the plurality of image data sets:

detecting an object in the image by performing foreground detection on the image to determine contours of the object;

determining feature descriptors of the image;

classifying the object in accordance with an object category;

determining one or more attributes of the object, based on the feature descriptors within the contours of the object; and

generating a visual token associated with the image, the visual token comprising information including the object category and the one or more attributes of the object determined from the feature descriptors;

store the visual tokens in the token database;

receive a trend query relating to an object category and defining two or more groups of interest for which trend data is to be determined with a defined time period;

identify visual tokens of interest relating to the object category for use in searching the token database;

identify one or more image data sets associated with each of the two or more groups of interest;

analyze visual tokens matching the visual tokens of interest that are associated with the identified image data sets to generate trend data for each of the two or more groups of interest for the defined period of time; and

generate an output based on a comparison of the trend data generated for the two or more groups of interest for the defined period of time.

17. The system of claim 16 , wherein the trend data generated by analyzing the visual tokens is indicative of a presence of an object attribute in relation to the object category over time.

18. The system of claim 16 , wherein the at least one processor is configured to display the output on a display.

19. The system of claim 16 , wherein the at least one processor is configured to generate an activation for the trend, the activation comprising one or more scheduled activities relating to the trend data.