WorldComp2D: Spatio-semantic Representations of Object Identity and Location from Local Views

Open paper

Benchmark	Model	Rank	Results
face-alignment-on-300w	WorldComp2D	–	NME_inter-ocular (%, Full): 5.06
face-alignment-on-cofw	WorldComp2D	–	NME (inter-ocular): 5.16%