Monocular depth estimation

Train DNN with massive pairs of image + depth map. There are many open codes, but a few codes share the weights (~ training with GPU is not necessary).

One example is
I. Alhashim and P. Wonka, “High Quality Monocular Depth Estimation via Transfer Learning,” arXiv:1812.11941

https://github.com/ialhashim/DenseDepth

 

See “Demos”
(https://github.com/ialhashim/DenseDepth/blob/master/DenseDepth.ipynb is good example)

> git clone https://github.com/ialhashim/DenseDepth.git

> wget https://s3-eu-west-1.amazonaws.com/densedepth/nyu.h5 -O ./DenseDepth/nyu.h5

> python test.py –input input/*.bmp

 

(My experience:)

For Anaconda:

> conda install keras

> conda install tensorflow

> conda install scikit-image

Add code on utils.py (to solve OpenMP error #15):

import os

os.environ[‘KMP_DUPLICATE_LIB_OK’] = ‘TRUE’