Skip to content

Latest commit

 

History

History
10 lines (7 loc) · 430 Bytes

README.md

File metadata and controls

10 lines (7 loc) · 430 Bytes

Scene-Graph-Generation

Prediction of action and spatial visual relationships in images between objects in the VRD-Dataset. It works in six phases: 1). Date pre-processing and augmentation using clustering 2). Object detection through which buildinga fully connected graph 3). Features generation- visual, semantic, spatial 4). Heatmap generation forevery relationship 5). Predicate detection using structural ranking loss