search

vincentschen / limited-label-scene-graphs

Scene Graph Prediction with Limited Labels
MIT License
54 stars 5 forks source link

Not able to get predicate form image #5

Open rohi0007 opened 3 years ago

rohi0007 commented 3 years ago

Hi Sir, I am a student and very inspired from your work.

For learning, I want to train model on some more object and predicates like image including "tree" and "house" and predicate is "hanging over" and the use case which I am trying to solve is to identify if a tree is hanging over house roof or not.

I understood the flow of main.ipynb and executed properly after downloaded all required data sets. but at the end also, did't get predicates from object only like "person sidewalk" this is coming where only subject and object is coming predicate is not coming.

requesting you please help to understand, how can we solve this.

I also tried to follow your tutorial - https://github.com/snorkel-team/snorkel-tutorials/tree/master/visual_relation I also execute this successfully and created .pth model but not able to understand how can I use this created model for predicting, subject/object and predicate from a input image.

Please help into this, I am stuck with this from a long time.

supriamir commented 2 years ago

@rohi0007 did you get the answer to use the model and show the image output with their relationship?