Skip to main content
Springer Nature Link
Account
Fig. 10 | Journal of King Saud University Computer and Information Sciences

Fig. 10

From: AMCA-SGG: Adaptive multi-prototype learning and cross-modal alignment for unbiased scene graph generation

Fig. 10

Qualitative comparison of scene graph generation results on the PredCls task. The left column shows the predicted relation graphs of the VCTree+GCL, and the right column presents the results of our proposed VCTree+AMCA method. Blue edges denote correctly predicted predicates, red solid edges denote incorrectly predicted predicates, and purple edges denote fine-grained predicate predictions

Back to article page