Learning multiple visual domains with residual adapters

Open in new window