Post #90

​Scene Graph Generation by Iterative Message Passing There are numerous applications of graphs in computer vision, which usually boil down to reduction of the image to a sparse graph. This has found applications in image captioning, human action identification, predicting trajectories, or transferring knowledge to new data sets, among others (see Applications section from this survey). One of the recent interesting advancements is from 2017 CVPR paper (a group of Fei-Fei Li), Scene Graph Generation by Iterative Message Passing, where the authors proposed an image encoder that produces essentially a knowledge graph of objects in the image and their relationships, which can be then used for downstream tasks.