System and method for image processing for trend forecasting

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

providing a database storing a plurality of visual tokens, 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;

generating, using at least one processor, visual tokens associated with images from a plurality of image data sets, wherein generating the visual tokens using the at least one processor 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 the database in association with an identifier that identifies the visual token as being associated with an image data set;

determining, using the at least one processor, trendsetter visual tokens of interest associated with an image data set of one or more trendsetters for a particular object category;

determining, using the at least one processor, a trendsetter adoption of one or more particular attributes of an object for the particular object category from the trendsetter visual tokens of interest;

determining, using the at least one processor, mainstream visual tokens of interest associated with an image data set of one or more mainstream adopters for the particular object category;

determining, using the at least one processor, a mainstream adoption of the one or more particular attributes of the object for the particular object category from the mainstream visual tokens of interest; and

determining, using the at least one processor, a state of a trend for the one or more particular attributes of the object based on the trendsetter adoption and the mainstream adoption.

2 . The method of claim 1 , wherein:

determining the trendsetter adoption comprises determining a percentage of trendsetter visual tokens having the one or more particular attributes of the object among the trendsetter visual tokens of interest; and

determining the mainstream adoption comprises determining a percentage of mainstream visual tokens having the one or more particular attributes of the object among the mainstream visual tokens of interest.

3 . The method of claim 2 , wherein:

determining the trendsetter adoption further comprises determining whether the percentage of trendsetter visual tokens having the one or more particular attributes of the object among the trendsetter visual tokens of interest is increasing or decreasing over a pre- defined period of time; and

determining the mainstream adoption further comprises determining whether the percentage of mainstream visual tokens having the one or more particular attributes of the object among the mainstream visual tokens of interest is increasing or decreasing over the pre-defined period of time.

4 . The method of claim 1 , further comprising forecasting a future state of the trend based on the state of the trend.

5 . The method of claim 4 , wherein forecasting the future state of the trend comprises determining a trend growth potential based on the trendsetter adoption and the mainstream adoption.

6 . The method of claim 4 , wherein forecasting the future state of the trend comprises determining an abandonment rate for the trend.

7 . The method of claim 1 , further comprising determining the one or more trendsetters for the particular object category by:

reviewing historical trend data to determine one or more previous trend attributes of an object for the particular object category, and a date of mainstream adoption of the one or more previous trend attributes of the object;

determining historical visual tokens of interest for the particular object category before the date of mainstream adoption; and

determining trendsetters for the object category from the historical visual tokens of interest comprising the one or more previous trend attributes of the object.

8 . The method of claim 7 , further comprising determining sub-categories of the trendsetters based on a date of the historical visual tokens comprising the one or more previous trend attributes of the object before the date of mainstream adoption.

9 . The method of claim 8 , further comprising determining the state of the trend based on the trendsetter adoption of sub-categories of trendsetters.

10 . The method of claim 7 , wherein the historical trend data comprises market data over time for the one or more previous trend attributes of the object.

11 . The method of claim 7 , wherein the mainstream adopters for the object category do not have associated historical visual tokens before the date of mainstream adoption of the historical trend data.

12 . The method of claim 11 , further comprising calibrating the mainstream adopters for the object category by correlating mainstream visual tokens having the one or more particular attributes of the object with retail sales data.

13 . The method of claim 1 , wherein the plurality of image data sets are associated with a plurality of social media accounts.

14 . 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.

15 . A system for forecasting trends using image data, comprising:

a database storing a plurality of visual tokens, 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; and

a server comprising at least one processor configured to:

generate 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;

store the visual tokens in a 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;

determine trendsetter visual tokens of interest associated with an image data set of one or more trendsetters for a particular object category;

determine a trendsetter adoption of one or more particular attributes of an object for the particular object category from the trendsetter visual tokens of interest;

determine mainstream visual tokens of interest associated with an image data set of one or more mainstream adopters for the particular object category;

determine a mainstream adoption of the one or more particular attributes of the object for the particular object category from the mainstream visual tokens of interest; and

determine a state of a trend for the one or more particular attributes of the object based on the trendsetter adoption and the mainstream adoption.

16 . The system of claim 15 , wherein:

determining the trendsetter adoption comprises determining a percentage of trendsetter visual tokens having the one or more particular attributes of the object among the trendsetter visual tokens of interest; and

determining the mainstream adoption comprises determining a percentage of mainstream visual tokens having the one or more particular attributes of the object among the mainstream visual tokens of interest.

17 . The system of claim 16 , wherein:

determining the trendsetter adoption further comprises determining whether the percentage of trendsetter visual tokens having the one or more particular attributes of the object among the trendsetter visual tokens of interest is increasing or decreasing over a pre- defined period of time; and

determining the mainstream adoption further comprises determining whether the percentage of mainstream visual tokens having the one or more particular attributes of the object among the mainstream visual tokens of interest is increasing or decreasing over a pre-defined period of time.

18 . The system of claim 15 , wherein the at least one processor is further configured to forecast a future state of the trend based on the current state of the trend.

19 . The system of claim 18 , wherein forecasting the future state of the trend comprises determining a trend growth potential based on the trendsetter adoption and the mainstream adoption, and/or determining an abandonment rate for the trend.

20 . The system of claim 15 , wherein the at least one processor is further configured to output the state of the trend.