CVPR 2020 release
Complete set of features described in our CVPR 2020 oral: 3D Packing for Self-Supervised Monocular Depth Estimation.
Additions from v0.1.0:
- Velocity loss (see VelSupModel)
- PackNetSlim01 (faster version of PackNet)
- Support for multi-camera loading on the DGP dataset
- Support for fp16 at inference time