Generating kin images using Generative Adversarial Networks
Kin-images generated by FamilyGAN. Comparing results from two different versions of our model.
Establising the problem statement on generating images of kin (family members). FamilyGAN achieves ~2% variation in kin-verification scores for original and generated pairs. This indicates that the generated distribution successfully observes kin feature hierarchies.
Raunak Sinha
Staff Research Engineer, Artificial Intelligence
I am interested in developing learning theory for computational sustainability, computer vision and natural language understanding.