search

microsoft / scene_graph_benchmark

image scene graph generation benchmark
MIT License
382 stars 86 forks source link

VinVL can model the relation prediction? #22

Closed alice-cool closed 3 years ago

alice-cool commented 3 years ago

I found VinVL 'S object and attribute lable is so bigger. So How to use the VinVL in predicate classification. At Now the project only provides 150 object. But visual genome +faster rcnn can detect 1370 object class. It is so big difference.

alice-cool commented 3 years ago

sgcls maybe not right. The code running failed

out_dists_sampled[is_overlap[box_ind, :, cls_ind], cls_ind] = 0.0 IndexError: index 61 is out of bounds for axis 2 with size 1 Killing subprocess 24877

alice-cool commented 3 years ago

if we don't provide relations we will get the following "exception" AttributeError: 'SceneParser' object has no attribute 'relation_head'

alice-cool commented 3 years ago

image So the flexible label provides whether it only can constrained to the predifined 151 object label?

alice-cool commented 3 years ago

Dear scholar, Could the VinVL model can produce specific features for my designated or the grounding truth coordinate?

hanxiaotian commented 3 years ago

For releation predictions, you need to run our pretrained relation prediction models. You can refer to this https://github.com/microsoft/scene_graph_benchmark/issues/16#issuecomment-876623915