Methods and machines for generating depth information

Abstract

Systems and techniques are provided to generate depth information. For example, the process might include a step to acquire a first feature volume comprising visual features corresponding to each respective frame in the first set of frames. A first query generation network can generate reconstructed features linked to the reconstructed feature volume corresponding to the first feature volume. Based on the first feature volume, a second query generation network can generate motion features linked to the predicted future motion corresponding to the first feature volume. Initial depth predictions can be generated for each respective frame based on cross-referencing between the features of the depth prediction decoder, reconstructed features, and motion features. Refined depth predictions can be generated for each respective object based on cross-referencing between the initial depth prediction, reconstructed features, and motion features.

Images