RobertoIV / FCRNN

Monocular depth estimation with FCRN plus surface normal constraint

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

FCRNN

Monocular depth estimation with FCRN from Deeper Depth Prediction with Fully Convolutional Residual Networks with internal surface normal constraint. Note this project is experimental and not carefully organized.

Requirement

  • Python 3.6
  • PyTorch 0.4.0
  • TensorboardX
  • Other common packages like numpy, skimage

Training

Download NYU_DEPTH v2 and train/test splits matrix into data. Run python run.py.

Demo

Download pretrained model from here and put it in checkpoints/normal_internal. Then run demo.ipynb using jupyter.

About

Monocular depth estimation with FCRN plus surface normal constraint

License:MIT License


Languages

Language:Jupyter Notebook 90.3%Language:Python 9.7%