Supplementary materials DADA: Depth-aware Domain ...openaccess.thecvf.com/.../Vu_DADA...supplemental.pdfSupplementary materials DADA: Depth-aware Domain Adaptation in Semantic Segmentation

Supplementary materialsDADA: Depth-aware Domain Adaptation in Semantic Segmentation

Tuan-Hung Vu1 Himalaya Jain1 Maxime Bucher1 Matthieu Cord1,2 Patrick Perez1

1valeo.ai, Paris, France 2Sorbonne University, Paris, France

(a) Input image (b) GT (c) AdvEnt (d) DADA

Additional qualitative results in the SYNTHIA→Cityscapes (16 classes) set-up. The four columns plot (a) RGB input images, (b)ground-truths, (c) AdvEnt baseline outputs and (d) DADA predictions. DADA shows good performance on ‘bus’, ‘car’, ‘bicycle’ classes.Best viewed in color.

1