Code for Faster rcnn

[shunlinlu]

Is it possible to release the faster rcnn part for image embedding?

[Gazeal]

I am sorry we currently can not share that source code. However, you can follow this work (https://github.com/bknyaz/sgg) to have a Visual Genome pretrained model, which can use to extract scenegraphs (objects with relations) for input images. For objects with attributes, You can use this work (https://github.com/peteanderson80/bottom-up-attention). The mentioned extractions are easy to reformat into predicates.