run_pipeline.sh

#!/bin/bash

# run this script under the directory of monodepth2
# you may need to first put all the test images in a test_images folder under assets/
# you will get the depth output for monodepth presentation, o3d and infiniTAM

function reconstruct_o3d() {
  # move scene image and o3d depth to o3d reconstruction pipeline
  mkdir $dir/Reconstruction_o3d/ReconstructionSystem/dataset/kitti_1
  cp -r $dir/DepthPrediction/assets/test_images/ $dir/Reconstruction_o3d/ReconstructionSystem/dataset/kitti_1/image/
  cp -r $dir/DepthPrediction/assets/output_depth_o3d/ $dir/Reconstruction_o3d/ReconstructionSystem/dataset/kitti_1/depth/

  echo "===================================="
  echo "Start Reconstruction Using Open3d"
  echo "===================================="

  cd $dir/Reconstruction_o3d/ReconstructionSystem/ || exit
  python run_system.py $dir/Reconstruction_o3d/ReconstructionSystem/config/kitti_1.json --make --register --refine --integrate

  echo "===================================="
  echo "Visualize Reconstruction"
  echo "===================================="
  cd $dir/src || exit
  python pcd_vis.py --scene_path $dir/Reconstruction_o3d/ReconstructionSystem/dataset/kitti_1/scene/integrated.ply --ext ply

}

function reconstruct_infinitam() {

  # first need to convert rgb to ppm format and depth to pgm format
  echo "===================================="
  echo "Convert depth and rgb image to infinitam format"
  echo "===================================="
  cp -r $dir/DepthPrediction/assets/test_images/ $dir/DepthPrediction/assets/output_rgb_infinitam/

  cd $dir/DepthPrediction/assets/output_rgb_infinitam/ && magick mogrify -format ppm *.png
  cd $dir/DepthPrediction/assets/output_depth_infinitam/ && magick mogrify -format pgm *.png

  echo "===================================="
  echo "Done"
  echo "===================================="

  # run InfiniTAM
  echo "===================================="
  echo "Start Reconstruction Using InfiniTAM"
  echo "===================================="

  # shellcheck disable=SC2164
  cd $dir/Reconstruction/InfiniTAM/build/Apps/InfiniTAM/
  ./InfiniTAM $dir/Reconstruction/InfiniTAM/kitti/calib3.txt \
    $dir/DepthPrediction/assets/output_rgb_infinitam/%04i.ppm \
    $dir/DepthPrediction/assets/output_depth_infinitam/%04i.pgm

  echo "===================================="
  echo "Reconstruction Finished"
  echo "===================================="

  echo "===================================="
  echo "Visualize Reconstruction"
  echo "===================================="

  cd $dir/src || exit
  python pcd_vis.py --scene_path $dir/Reconstruction/InfiniTAM/build/Apps/InfiniTAM/color_mesh.obj --ext obj
}

function auto_ar_followup() {

  echo "===================================="
  echo "AUTO MODE FOR AR"
  echo "===================================="
  # shellcheck disable=SC2164
  cd $dir/AR/
  blender -b --python auto_follow_up.py --\
  --obj_dir $dir/AR/scaled_objs \
  --obj "Bus" \
  --bg $dir/AR/subscenes/ \
  --env $dir/AR/hdrs/ \
  --out $dir/AR/outputs/ \
  --pose $dir/AR/poses/ \
  --light $dir/AR/light.txt
}

echo "===================================="
echo "Set up folders for specific outputs"
echo "===================================="

export dir=$PWD

mkdir $dir/DepthPrediction/assets/output_depth_rendering/
mkdir $dir/DepthPrediction/assets/output_depth_infinitam/
mkdir $dir/DepthPrediction/assets/output_npy_mono/
mkdir $dir/DepthPrediction/assets/output_depth_o3d/

## Rename the name of the image from 10-digits to 4
## Only need to run ONCE for a new dataset
#for file in ./DepthPrediction/assets/test_images/*
#do
#  lastfour=${file:23}
##  lastfour="${file:27:30}"
#  filename=${file:0:21}$lastfour
#  echo $filename
#
#  mv $file $filename
#done

echo "===================================="
echo "Render input rgb images to video"
echo "===================================="
cd $dir/src || exit
python pic2video.py --image_path $dir/DepthPrediction/assets/test_images/ --output_name input.mp4
echo "===================================="
echo "See rendered video under src directory"
echo "===================================="

echo "===================================="
echo "Depth Prediction Started"
echo "===================================="

for file in $dir/DepthPrediction/assets/test_images/*; do
  cd $dir/DepthPrediction/ || exit
  python test_simple.py --image_path $file --output_depth $dir/DepthPrediction/assets/output_depth_rendering/ \
    --output_npy $dir/DepthPrediction/assets/output_npy_mono/ \
    --output_depth_o3d $dir/DepthPrediction/assets/output_depth_o3d/ \
    --output_depth_infi $dir/DepthPrediction/assets/output_depth_infinitam/ \
    --model_name mono_1024x320 \
    --no_cuda
done

echo "===================================="
echo "Depth Prediction Finished"
echo "===================================="

echo "===================================="
echo "Render output depth images to video"
echo "===================================="
cd $dir/src || exit
python pic2video.py --image_path $dir/DepthPrediction/assets/output_depth_rendering/ --output_name depth.mp4
echo "===================================="
echo "See rendered video under src directory"
echo "===================================="

read -r -p "Do you want to reconstruct using Open3d? [y/N] " response
if [[ "$response" =~ ^([yY][eE][sS]|[yY])$ ]]; then
  reconstruct_o3d
fi

read -r -p "Do you want to reconstruct using InfiniTAM? [y/N] " response
if [[ "$response" =~ ^([yY][eE][sS]|[yY])$ ]]; then
  reconstruct_infinitam
fi

read -r -p "Do you want to use Blender script to automatically generate AR images? [y/N] " response
if [[ "$response" =~ ^([yY][eE][sS]|[yY])$ ]]; then
  auto_ar_followup
fi

wait

echo "===================================="
echo "All Done"
echo "===================================="