US 12,353,960 B2
Engagement prediction of image assets on social media
Jacek Wolkowicz, Halifax (CA); and Tomasz Niewiarowski, Halifax (CA)
Assigned to Dash Hudson, Halifax (CA)
Filed by Dash Hudson, Halifax (CA)
Filed on Nov. 26, 2019, as Appl. No. 16/696,569.
Prior Publication US 2021/0158208 A1, May 27, 2021
Int. Cl. G06N 20/00 (2019.01); G06F 16/9535 (2019.01); G06N 3/045 (2023.01); G06N 3/08 (2023.01); G06Q 50/00 (2024.01)
CPC G06N 20/00 (2019.01) [G06F 16/9535 (2019.01); G06N 3/045 (2023.01); G06N 3/08 (2013.01); G06Q 50/01 (2013.01)] 28 Claims
OG exemplary drawing
 
1. A non-transitory computer readable medium comprising instructions stored thereon, the instructions being effective to cause at least one processor to:
receive a collection of candidate images that are candidates for posting on a social media platform;
convert images of the collection of candidate images into a high-dimensional feature vector representing content of images of the collection of candidate images, using a trained deep learning model, wherein the feature vector is generated by an image conversion service;
determine, using at least one artificial intelligence model including a brand model and a network model, a prediction for the images of the collection of candidate images, wherein the prediction includes an expected engagement on the social media platform, the expected engagement being a function of prior interactions with images containing similar feature vectors;
receive an output from the brand model where the output from the brand model includes a brand-specific output engagement rate for the collection of candidate images, wherein the brand-specific output engagement rate is determined based on a first number of interactions with the collection of candidate images on the social media platform related to the brand model;
receive an output from the network model where the output from the network model includes a general output engagement rate for the collection of candidate images, wherein the general output engagement rate is determined based on a second number of interactions with the collection of candidate images on the social media platform not related to the brand model;
combine the brand-specific output engagement rate and the general output engagement rate to generate a combined engagement rate prediction for the collection of candidate images representative of the expected engagement on the social media platform; and
output, for display, a set of candidate images of the collection of candidate images into one or more groups based on one or more similarities of the combined engagement rate prediction of images in the set of candidate images.