No adaptation head: discriminate on features

[vict0rsch]

Instead of using A to structure the latent space and make it domain-invariant (in terms of domain adaptation) we can use an output discriminator after S and D for instance to make those representations indiscernable real | sim

[vict0rsch]

1. Learning to Adapt Structured Output Space for Semantic Segmentation: https://arxiv.org/pdf/1802.10349.pdf
2. ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation: https://arxiv.org/pdf/1811.12833.pdf
3. Domain Adaptation for Structured Output via Discriminative Patch Representations: https://arxiv.org/pdf/1901.05427v3.pdf
4. DADA: Depth-aware Domain Adaptation in Semantic Segmentation: https://arxiv.org/abs/1904.01886

[vict0rsch]

https://paperswithcode.com/sota/image-to-image-translation-on-synthia-to

State-of-the-art table for Image-to-Image Translation on SYNTHIA-to-Cityscapes

A performance comparison of 7 methods.