Deep Structure from Motion

My undergraduate neural nets class final project

My team and I took results and code from a Berkeley and Google research group and improved their baseline depth map accuracies by incorporating loss from a Mask-RCNN.

Link to poster: poster

Link to short paper: paper