System and method for providing similar or related products based on deep-learning

A method for detecting similar or related products through deep-learning, which is performed by a processor of a shopping mall server, and transmitting the detected similar or related products to a terminal of a user, the method comprising:

• acquiring an image of an item selected by the user and item information of the item selected by the user;
• detecting one or more objects associated with the item in the image and generating one or more bounding boxes for the detected one or more objects;
• determining a search target item based on the detected one or more objects;
• cropping a portion of the image disposed in a bounding box of an object associated with the search target item, and performing padding-processing to the cropped portion of the image disposed in the bounding box of the object associated with the search target item;
• extracting a feature vector by deep-learning the cropped and padding-processed portion of the image;
• and
• detecting an item among items stored in a database based on a similarity between the feature vector extracted by the deep-learning and feature vectors of the items stored in the database and transmit the detected item to the terminal of the user.

An apparatus for providing one or more similar or related products using deep-learning, the apparatus comprising:

• a communication unit configured to transmit and receive data associated with a shopping service to and from a shopping mall server;
• a storage unit configured to store the data associated with the shopping service;
• a display unit configured to output the data associated with the shopping service including an image of a product selected by a user;
• and
• a processor configured to provide one or more images of one or more similar or related products associated with the image of the product selected by the user,
• wherein the processor is configured to:
• acquire the image of the product selected by the user and product information of the product selected by the user,
• detect one or more objects in the image of the product and create one or more bounding boxes for the detected one or more objects,
• determine a search target product based on the detected one or more objects,
• crop a portion of the image disposed in a bounding box of an object associated with the search target product, and perform padding-processing to the cropped portion of the image disposed in the bounding box of the object associated with the search target item,
• extract a feature vector by deep-learning the cropped and padding-processed portion of the image,
• detect a product among products stored in the database based on a similarity between the feature vector extracted by the deep-learning and feature vectors of the products stored in the database, and
• output the detected product through the display unit.