I would like to start implementing selective compression so that the network can be instructed to compress/synthesize certain regions while preserving selected regions. (Section 4.2 of https://arxiv.org/pdf/1804.02958.pdf). This will require working with a semantic map of the original image, which is available for the Cityscapes/ ADE20k dataset.
I would like to start implementing selective compression so that the network can be instructed to compress/synthesize certain regions while preserving selected regions. (Section 4.2 of https://arxiv.org/pdf/1804.02958.pdf). This will require working with a semantic map of the original image, which is available for the Cityscapes/ ADE20k dataset.