Multi-task Learning Using Multi-modal Encoder-Decoder Networks with Shared Skip Connections

Multi-task learning is a promising approach for efficiently and effectively addressing multiple mutually related recognition tasks. Many scene understanding tasks such as semantic segmentation and depth prediction can be framed as cross-modal encoding/decoding, and hence most of the prior work used multi-modal datasets for multi-task learning. However, the inter-modal commonalities, such as one across image, depth, and semantic labels, have not been fully exploited. We propose a multi-modal encoder-decoder networks to harness the multi-modal nature of multi-task scene recognition. In addition to the shared latent representation among encoder-decoder pairs, our model also has shared skip connections from different encoders. By combining these two representation sharing mechanisms, the proposed method efficiently learns a shared feature representation among all modalities in the training data. Experiments using two public datasets shows the advantage of our method over baseline methods that are based on encoder-decoder networks and multi-modal auto-encoders.