Method and system of advertising

A system of advertising, the system comprising:

• a camera, adapted for recording a source video;
• a memory unit, adapted for storing the source video;
• a shot segmentation module comprising a first CPU;
• an object detection module comprising a second CPU and a first GPU;
• an advertisement-image retrieval module comprising a third CPU and a second GPU;
• an advertisement-target-object optimization module comprising a fourth CPU;
• and
• an advertisement-insertion-position optimization module comprising a fifth CPU;
• wherein:
• the system is connected to a plurality of displays;
• the first CPU is adapted to calculate and segment the source video into individual scenes based on clustering-based approach, and to store the individual scenes in the memory unit;
• the first GPU is adapted to calculate and obtain information about objects in the source video for each individual scene by region-wise convolutional characteristics based detection, and to store the information about the objects in the source video in the memory unit;
• the second GPU is adapted to search in a database for advertisement objects that match the objects using garment retrieval and a category-based strategy;
• the fourth CPU is adapted to perform optimization processing of the retrieved advertisement objects that match the objects in the source video to obtain a candidate advertisement;
• and
• the fifth CPU is adapted to add the candidate advertisement into the source video, to optimize the distance between the candidate advertisement and a target object and the area of overlapping regions between the candidate advertisement and the objects in the source video to obtain the video that comprises the candidate advertisement, to distribute the video that comprises the candidate advertisement to the plurality of displays, and to display the video that comprises the candidate advertisement on the plurality of displays.