System and method for multi attribute based data synthesis

1 . A system for creating a multi-attribute-based data synthesis, said system comprising:

one or more processors; and

a memory operatively coupled to the one or more processors, wherein the memory comprises processor-executable instructions, which on execution, cause the one or more processors to:

obtain an input image from a database, wherein the image comprises a foreground and a background, and wherein the input image obtained from the database comprises an image of a human wearing an apparel in the foreground and a scene in the background;

perform a coarse segmentation of the input image to obtain a set of mask labels;

annotate the set of mask labels to differentiate the foreground from the background;

apply a weakly supervised fine-segmentation algorithm, wherein an annotator corrects segmentation labels by clicking and adding dots to the foreground and crosses to the background, and a graph cut algorithm utilizes the corrections to re-segment the image;

create a plurality of images by varying the annotated set of mask labels;

vary one or more parameters associated with the human in the input image; and

for each of the plurality of images, vary placement of the foreground with respect to the background, wherein the image of the human with varying apparel is placed at different, non-overlapping spatial locations within a single, fixed background scene.

2 . The system as claimed in claim 1 , wherein the memory comprises processor-executable instructions, which on execution, cause the one or more processors to:

detect a face region of the human in the image;

determine a top and torso location of the human in the image with respect to scene in the background; and

generate a set of priors for creating the plurality of images.

3 . The system as claimed in claim 1 , wherein the memory comprises processor-executable instructions, which on execution, cause the one or more processors to:

randomize one or more parameters of the apparel in the input image.

4 . A method for creating a multi-attribute-based data synthesis, said method comprising:

obtaining, by one or more processors, an input image from a database, wherein the input image obtained from the database comprises a human wearing an apparel in a foreground and a scene in a background;

detecting, by the one or more processors, the human in the input image;

detecting, by the one or more processors, a face region of the human in the input image;

generating, by the one or more processors, a body-ratio-heuristics for the input image;

generating, by the one or more processors, a set of mask labels associated with the input image;

performing, by the one or more processors, a domain randomization of the input image to generate a plurality of images to create the image database;

varying one or more parameters associated with the human in the input image; and

for each of the plurality of images, varying placement of the foreground with respect to the background, wherein the image of the human with varying apparel is placed at different, non-overlapping spatial locations within a single, fixed background scene.

5 . The method as claimed in claim 4 , wherein generating, by the one or more processors, the body-ratio-heuristics comprises predicting, by the one or more processors, a position of torso, top, and bottom of the human in the input image with respect to the scene in the background.

6 . The method as claimed in claim 4 , wherein the set of mask labels comprise at least one of: a sure foreground, a sure background, a probable foreground, and a probable background.

7 . The method as claimed in claim 4 , comprising:

randomizing one or more parameters of the apparel in the input image.