1 code implementation • NeurIPS 2021 • Sunyong Seo, Hana Kim, Jinho Park
Finally, we suggest a deformable loss function by the composition of the predicted image through the model to avoid the one-to-many problems of the dataset.