Add docker scripts to gaussian-splatting

docker/Dockerfile

@@ -0,0 +1,84 @@
+#FROM tensorflow/tensorflow:latest-gpu
+FROM nvidia/cuda:12.2.0-devel-ubuntu22.04
+
+ENV DEBIAN_FRONTEND noninteractive
+
+ARG user_id
+ARG root_psw="12345678"
+ARG user_psw="ok"
+ARG user_name=user
+
+# Installs the necessary pkgs.
+RUN \
+	echo "**** packages installation ****" \
+		&& apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/3bf863cc.pub \
+		&& apt-get update && apt-get dist-upgrade -y && apt-get install -y \
+			vim \
+            build-essential \
+            cmake \
+            ffmpeg \
+            imagemagick \
+            unzip \
+            freeglut3-dev \
+            libopencv-dev \
+            libjpeg-dev \
+            libpng-dev \
+            libglew-dev \
+            libpthread-stubs0-dev \
+			git \
+			virtualenv \
+            sqlite3 \
+            libsqlite3-dev \
+			time \
+			sudo \
+			wget \
+			nano \
+            libboost-program-options-dev \
+            libboost-filesystem-dev \
+            libboost-graph-dev \
+            libboost-regex-dev \
+            libboost-system-dev \
+            libboost-test-dev \
+            libeigen3-dev \
+            libsuitesparse-dev \
+            libfreeimage-dev \
+            libgoogle-glog-dev \
+            libgflags-dev \
+            libglew-dev \
+            qtbase5-dev \
+            libqt5opengl5-dev \
+            libcgal-dev \
+            libcgal-qt5-dev \
+            libmetis-dev \
+            libflann-dev \
+            libatlas-base-dev \
+            libsuitesparse-dev \
+            libcufft10 \
+            libcusparse11 \
+            libcublas11 \
+            libcublaslt11 \
+	&& echo "**** python pip update ****" \
+		&& /usr/bin/python3 -m pip install --upgrade pip \
+	&& echo "**** aliases for l and ll commands creation ****" \
+		&& echo -e 'ls --color=auto "$@"' > /usr/bin/l \
+		&& echo -e 'ls -lah --color=auto "$@"' > /usr/bin/ll \
+    	&& chmod +x /usr/bin/ll /usr/bin/l \
+	&& echo "**** history-search-backward by pressing F8 ****" \
+		&& sed -i 's/# "\\e\[5~": history-search-backward/"\\e\[19~": history-search-backward/' /etc/inputrc \
+	&& echo "**** root password creation ****" \
+		&& echo "root:${root_psw}" | chpasswd \
+	&& echo "**** user creation ****" \
+		&& useradd -m -s /usr/bin/bash -u ${user_id} -G sudo ${user_name} \
+		&& echo "${user_name}:${user_psw}" | chpasswd \
+		&& chown -R ${user_name}:${user_name} /home/${user_name}/ \
+		&& mkdir /home/${user_name}/workspace/ \
+        && chown -R user:user /home/${user_name}/workspace
+
+
+USER ${user_name} 
+WORKDIR /home/${user_name}/
+
+
+RUN git config --global http.sslverify false
+RUN cd /home/${user_name}/workspace && git clone --recursive https://github.com/graphdeco-inria/gaussian-splatting && cd gaussian-splatting && docker/initialize.sh
+

docker/build_and_deploy.sh

@@ -0,0 +1,44 @@
+#!/usr/bin/env bash
+# This script builds and runs a docker image for local use.
+
+#Although I dislike the use of docker for a myriad of reasons, due needing it to deploy on a particular machine
+#I am adding a docker container builder for the repository to automate the process
+
+
+DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
+cd "$DIR"
+cd ..
+REPOSITORY=`pwd`
+
+cd "$DIR"
+
+NAME="gaussian-splatting"
+dockerfile_pth="$DIR"
+mount_pth="$REPOSITORY"
+
+# update tensorflow image
+docker pull tensorflow/tensorflow:latest-gpu
+
+# build and run tensorflow
+docker build \
+	-t $NAME \
+	$dockerfile_pth \
+	--build-arg user_id=$UID
+
+docker run -d \
+	--gpus all \
+	--shm-size 8G \
+	-it \
+	--name $NAME-container \
+	-v $mount_pth:/home/user/workspace \
+	$NAME
+
+
+docker ps -a
+
+OUR_DOCKER_ID=`docker ps -a | grep gaussian-splatting | cut -f1 -d' '`
+echo "Our docker ID is : $OUR_DOCKER_ID"
+echo "Attaching it using : docker attach $OUD_DOCKER_ID"
+docker attach $OUR_DOCKER_ID
+
+exit 0

docker/initialize.sh

@@ -0,0 +1,59 @@
+#!/bin/bash
+
+
+#git clone https://github.com/NVIDIA/cuda-samples #<- To Test docker Cuda image..
+#cd cuda-samples
+
+
+#also get add-ons
+git clone https://github.com/antimatter15/splat
+git clone https://github.com/ReshotAI/gaussian-splatting-blender-addon/
+git clone https://github.com/francescofugazzi/3dgsconverter #needs scikit-learn
+
+
+
+
+git clone https://ceres-solver.googlesource.com/ceres-solver
+cd ceres-solver
+#git checkout $(git describe --tags) # Checkout the latest release
+mkdir build
+cd build
+cmake .. -DBUILD_TESTING=OFF -DBUILD_EXAMPLES=OFF -DCMAKE_CUDA_ARCHITECTURES="60;70;80" #-DUSE_CUDA=OFF 
+make -j8
+sudo make install
+cd ..
+cd ..
+
+
+git clone https://github.com/colmap/colmap
+cd colmap
+#git checkout dev
+mkdir build
+cd build
+cmake .. -DCMAKE_CUDA_ARCHITECTURES="60;70;80" #-DCUDA_ENABLED=OFF 
+make -j8
+sudo make install
+cd ..
+cd ..
+
+
+
+python3.10 -m pip install plyfile tqdm scikit-learn
+python3.10 -m pip install  https://huggingface.co/camenduru/gaussian-splatting/resolve/main/diff_gaussian_rasterization-0.0.0-cp310-cp310-linux_x86_64.whl
+python3.10 -m pip install  https://huggingface.co/camenduru/gaussian-splatting/resolve/main/simple_knn-0.0.0-cp310-cp310-linux_x86_64.whl
+python3.10 -m pip install torchvision
+
+ln -s docker/run.sh ./run.sh
+
+#Build viewer
+#sudo apt install -y libimgui-dev libglew-dev libassimp-dev libboost-all-dev libgtk-3-dev libopencv-dev libglfw3-dev libavdevice-dev libavcodec-dev libeigen3-dev libxxf86vm-dev libembree-dev
+#git clone https://github.com/JayFoxRox/SIBR_viewers
+#cd SIBR_viewers
+#cmake -Bbuild . -DCMAKE_BUILD_TYPE=Release -DASSIMP_LIBRARY=/usr/lib/x86_64-linux-gnu/libassimp.so
+#cmake --build build -j24 --target install
+#cd ..
+
+#sudo apt-get -y install cuda
+sudo apt -y install nvidia-cuda-toolkit
+
+exit 0

docker/run.sh

@@ -0,0 +1,33 @@
+#!/bin/bash
+
+DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
+cd "$DIR"
+
+
+#cd ..
+
+mkdir -p $1-data/input
+mkdir -p $1-data/output
+
+FPS="5" #<-change this to change framerate
+
+
+if [ -f $1-data/input/0001.jpg ]
+then
+  echo "File $1 appears to have already been split .."
+else
+  ffmpeg -i $1 -qscale:v 1 -qmin 1 -vf fps=$FPS $1-data/input/%04d.jpg
+fi
+
+
+python3.10 convert.py -s $1-data/ --camera SIMPLE_RADIAL --no_gpu #GPU produces worse results (?)
+python3.10 train.py -s $1-data/ -r 1 --model_path=$1-data/output/ --position_lr_init 0.000016 --scaling_lr 0.001 --iterations 35000 #Test more training budget
+
+python3.10 3dgsconverter/3dgsconverter.py -i $1-data/output/point_cloud/iteration_30000/point_cloud.ply -o $1-data/output/point_cloud/iteration_30000/output_cc.ply -f cc --rgb --density_filter --remove_flyers
+python3.10 3dgsconverter/3dgsconverter.py -i $1-data/output/point_cloud/iteration_30000/output_cc.ply  -o $1-data/output/point_cloud/iteration_30000/point_cloud_clean.ply -f 3dgs
+
+#pack it in
+#tar cvfjh "$1.tar.bz2" $1-data/
+
+
+exit 0

docker/setup.sh

@@ -0,0 +1,40 @@
+#!/usr/bin/env bash
+# This script builds and runs a docker image for local use.
+
+DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
+cd "$DIR"
+cd ..
+REPOSITORY=`pwd`
+
+cd "$DIR"
+
+#https://docs.nvidia.com/ai-enterprise/deployment-guide/dg-docker.html
+sudo apt-get update
+sudo apt-get install -y \
+    apt-transport-https \
+    ca-certificates \
+    curl \
+    gnupg-agent \
+    software-properties-common
+
+curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -
+sudo apt-key fingerprint 0EBFCD88
+
+sudo add-apt-repository \
+"deb [arch=amd64] https://download.docker.com/linux/ubuntu \
+$(lsb_release -cs) \
+stable"
+
+
+sudo apt-get update
+sudo apt-get install -y docker-ce docker-ce-cli containerd.io
+
+sudo docker run hello-world
+
+
+#Make sure docker group is ok
+sudo groupadd docker
+sudo usermod -aG docker $USER
+newgrp docker 
+
+exit 0

train.py

@@ -29,6 +29,10 @@
     TENSORBOARD_FOUND = False
 
 def training(dataset, opt, pipe, testing_iterations, saving_iterations, checkpoint_iterations, checkpoint, debug_from):
+    bestLossEncountered      = 1000000
+    bestIterationEncountered = 0
+    goodLossThreshold        = 0.07 
+
     first_iter = 0
     tb_writer = prepare_output_and_logger(dataset)
     gaussians = GaussianModel(dataset.sh_degree)
@@ -103,6 +107,16 @@ def training(dataset, opt, pipe, testing_iterations, saving_iterations, checkpoi
             if iteration == opt.iterations:
                 progress_bar.close()
 
+
+            if  (bestLossEncountered>ema_loss_for_log) and (goodLossThreshold > ema_loss_for_log):
+                     if (bestIterationEncountered!=0):
+                         print("\n[GOOD ITER {}] Erasing previous best iteration..".format(bestIterationEncountered))
+                         os.system("rm %s/point_cloud/iteration_%u/*.ply && rmdir %s/point_cloud/iteration_%u/"%(scene.model_path,bestIterationEncountered,scene.model_path,bestIterationEncountered))
+                     bestLossEncountered      = ema_loss_for_log
+                     bestIterationEncountered = iteration
+                     print("\n[GOOD ITER {}] Now remembering this iteration..".format(iteration))
+                     saving_iterations.append(iteration)
+
             # Log and save
             training_report(tb_writer, iteration, Ll1, loss, l1_loss, iter_start.elapsed_time(iter_end), testing_iterations, scene, render, (pipe, background))
             if (iteration in saving_iterations):