Open zyong812 opened 2 years ago
Hi @zyong812. I guess annotating is needed to create a training dataset. Almost all of the existing scene-graph generation models need to be trained with images having ground truth relationship labels (with object labels and/or attribute labels too), so hand labeling the image manually takes a lot of time. In such case, this sort of annotation software comes handy. I hope it answers your doubt. :)
Hi, I have also done some work on scene graph. I am just wondering why you annotate such dataset.