# Cloning the Repository 

!git clone https://github.com/pranjaldatta/DenseDepth-Pytorch.git

# Getting the data 
!python DenseDepth-Pytorch/densedepth/download_data.py

# Mounting drive
from google.colab import drive
drive.mount('/gdrive')


!mkdir /gdrive/My\ Drive/colabdrive/work/densedepth
!mkdir /gdrive/My\ Drive/colabdrive/work/densedepth/checkpoints

!pip install tensorboardX

# Prefer using Nvidia T4's or P100 for favourable training times


# memory footprint support libraries/code
!ln -sf /opt/bin/nvidia-smi /usr/bin/nvidia-smi
!pip install gputil
!pip install psutil
!pip install humanize
import psutil
import humanize
import os
import GPUtil as GPU
GPUs = GPU.getGPUs()
# XXX: only one GPU on Colab and isn’t guaranteed
gpu = GPUs[0]
def printm():
 process = psutil.Process(os.getpid())
 print("Gen RAM Free: " + humanize.naturalsize( psutil.virtual_memory().available ), " | Proc size: " + humanize.naturalsize( process.memory_info().rss))
 print("GPU RAM Free: {0:.0f}MB | Used: {1:.0f}MB | Util {2:3.0f}% | Total {3:.0f}MB".format(gpu.memoryFree, gpu.memoryUsed, gpu.memoryUtil*100, gpu.memoryTotal))
printm()

!nvidia-smi

!python DenseDepth-Pytorch/densedepth/train.py --epochs 10 \
                                              --data "data/nyu_depth.zip" \
                                              --batch 4 \
                                              --save "<path to save checkpoints in (prefer drive if using colab)" \
                                              --device "cuda" \
                                              --checkpoint "<path to checkpoint from which to resume training>"

!python DenseDepth-Pytorch/densedepth/test.py --checkpoint "<path to load weights from" \
                                              --device "cuda" \
                                              --data "DenseDepth-Pytorch/examples/"