SG-RoadSeg: End-to-End Collision-Free Space Detection Sharing Encoder Representations Jointly Learned via Unsupervised Deep Stereo | IEEE Conference Publication | IEEE Xplore