- Targeted on Monocular Depth Estimation based on NYU DepthV2 Dataset.
- Reimplemented the original Eigen paper (Torch/MATLAB) with PyTorch, also improved the performance by fusion of ResNet and FCN into Multi-Scale CNN training.
course project of Computational Photography Year 2018. June
- depth: We reproduce Eigen's Multi-Scale Paper (2014/5 NIPS) based on PyTorch(0.4).
- depth_fully_conv: Laina2016 FCNR
- check
presentation.pdf
Note: https://cs.nyu.edu/~deigen/depth/