Yaroslav Ganin

Papers

Total Citations

H-Index

About

Yaroslav Ganin is a leading researcher at the intersection of deep learning and 3D computer vision, with key contributions in generative modeling and domain adaptation. He is best known for pioneering work on adversarial domain adaptation, notably the "Domain-Adversarial Training of Neural Networks" (DANN), which introduced a gradient reversal layer to learn invariant features across domains—a foundational technique that has accrued thousands of citations and shaped modern transfer learning. In 3D generation, Ganin co-developed **PolyGen**, an autoregressive model that directly generates polygon meshes, a breakthrough for efficient 3D geometry representation in graphics and robotics. With over 65 citations on this work alone, PolyGen addresses the longstanding challenge of learning-based mesh generation, offering a principled alternative to voxel or point cloud approaches. His research, spanning adversarial learning, generative models, and 3D representation, has been published at top venues like NeurIPS, ICML, and CVPR, and his methods are widely adopted in both academia and industry. Ganin’s work continues to inspire new directions in robust, domain-adaptive AI and scalable 3D content creation.