Adding script to build Docker images with ETA installed

.dockerignore

@@ -0,0 +1 @@
+.git/modules

Dockerfile

@@ -0,0 +1,91 @@
+# Dockerfile for building a base image with an ETA environment atop a
+# Debian, Ubuntu, or related Linux distribution.
+#
+# ARGs:
+#   BASE_IMAGE
+#   TENSORFLOW_VERSION
+#
+# TensorFlow version notes:
+#   - For CPU-only images, use tensorflow==1.12.0
+#   - For GPU-enabled images, use the TensorFlow version compatible with the
+#     CUDA version you are using:
+#       - CUDA 8: tensorflow-gpu==1.4.0
+#       - CUDA 9: tensorflow-gpu==1.12.0
+#       - CUDA 10: tensorflow-gpu==1.14.0
+#
+# Copyright 2017-2020, Voxel51, Inc.
+# voxel51.com
+#
+# Brian Moore, [email protected]
+#
+
+#
+# The base image to build from, which must be a Debian, Ubuntu, or related
+# Linux distribution
+#
+
+ARG BASE_IMAGE
+FROM $BASE_IMAGE
+
+#
+# Install ETA
+#
+# Notes:
+#   ETA supports Python 2.7.X or Python 3.6.X
+#
+# `ppa:deadsnakes/ppa` is used in order to install Python 3.6 on Ubuntu 16.04
+# https://askubuntu.com/questions/865554/how-do-i-install-python-3-6-using-apt-get
+#
+# `https://bootstrap.pypa.io/get-pip.py` is used to install pip on Python 3.6
+# https://askubuntu.com/questions/889535/how-to-install-pip-for-python-3-6-on-ubuntu-16-10
+#
+# numpy==1.16.0 is enforced as a last step because tensorflow requires this
+# version to function properly, and some commands here seem to mess with the
+# numpy version installed via the `requirements.txt` file
+#
+
+RUN apt-get update \
+    && apt-get -y --no-install-recommends install software-properties-common \
+    && add-apt-repository -y ppa:deadsnakes/ppa \
+    && apt-get update \
+    && apt-get -y --no-install-recommends install \
+        sudo \
+        build-essential \
+        pkg-config \
+        ca-certificates \
+        cmake \
+        cmake-data \
+        unzip \
+        pciutils \
+        git \
+        curl \
+        wget \
+        python3.6 \
+        python3.6-dev \
+        libcupti-dev \
+        ffmpeg \
+        imagemagick \
+    && ln -s /usr/bin/python3.6 /usr/local/bin/python \
+    && curl https://bootstrap.pypa.io/get-pip.py | python
+
+WORKDIR /usr/src/app
+COPY . eta/
+
+ARG TENSORFLOW_VERSION
+RUN pip --no-cache-dir  install --upgrade pip setuptools \
+    && pip --no-cache-dir install -r eta/requirements.txt \
+    && pip --no-cache-dir install --upgrade setuptools \
+    && pip --no-cache-dir install -e eta/. \
+    && pip --no-cache-dir install -I $TENSORFLOW_VERSION \
+    && pip --no-cache-dir install --upgrade numpy==1.16.0 \
+    && curl -OL https://github.com/protocolbuffers/protobuf/releases/download/v3.6.1/protoc-3.6.1-linux-x86_64.zip \
+    && unzip protoc-3.6.1-linux-x86_64.zip -d protoc3 \
+    && rm -rf protoc-3.6.1-linux-x86_64.zip \
+    && mv protoc3/bin/* /usr/local/bin \
+    && mv protoc3/include/* /usr/local/include \
+    && rm -rf protoc3 \
+    && cd eta/tensorflow/models \
+    && protoc research/object_detection/protos/*.proto \
+        --proto_path=research \
+        --python_out=research \
+    && rm -rf /var/lib/apt

README.md

@@ -31,19 +31,28 @@ ETA is very portable:
 - Supports CUDA 8, 9 and 10 (for GPU installations)
 
 
-## Installation
+## Docker Installation
 
-0. _(Optional but highly recommended)_ Activate a
-[virtual environment](docs/virtualenv_guide.md)
+If you prefer to operate via Docker, see the
+[Docker Build Guide](docs/docker_build_guide.md) for simple instructions for
+building a Docker image with an ETA environment installed.
 
-1. Clone the repository:
+
+## Local Installation
+
+The following instructions describe how to install ETA locally on your machine.
+
+0. _(Optional but highly recommended)_
+[Activate a virtual environment](docs/virtualenv_guide.md)
+
+1. Clone the repository
 
 ```shell
 git clone https://github.com/voxel51/eta
 cd eta
 ```
 
-2. Run the install script:
+2. Run the install script
 
 ```shell
 bash install.bash
@@ -98,7 +107,7 @@ file to configure various package-level constants. Many advanced ETA features
 such as pipeline building, model management, etc. require a properly configured
 environment to function.
 
-To setup your environment, create a copy the example configuration file
+To setup your environment, create a copy the example configuration file:
 
 ```shell
 cp config-example.json config.json

docs/docker_build_guide.md

@@ -0,0 +1,93 @@
+# Docker Build Guide
+
+This document guides you through building a Docker image with a pre-installed
+ETA environment.
+
+
+## Dependencies
+
+- [Docker Community Edition](https://hub.docker.com/search/?type=edition&offering=community)
+- To build/run GPU-enabled images, you must install `nvidia-docker` by
+following [these instructions](https://github.com/NVIDIA/nvidia-docker)
+
+
+## Setup
+
+In order to build an ETA Docker image, you must have ETA cloned on your machine
+with a valid ETA `config.json` file installed:
+
+```shell
+# Clone repository
+git clone https://github.com/voxel51/eta
+cd eta
+
+# Clone submodules
+git submodule init
+git submodule update
+
+# Install your ETA config (customize as necessary)
+cp config-example.json config.json
+```
+
+
+## Building an image
+
+Follow the instructions below to build an ETA Docker image for your desired
+environment.
+
+These commands should be run from the root directory of your local ETA clone.
+
+
+#### CPU example
+
+The following snippet builds a CPU-enabled image on Ubuntu 16.04:
+
+```shell
+BASE_IMAGE=ubuntu:16.04
+TENSORFLOW_VERSION=tensorflow==1.12.0
+TAG=voxel51/eta:cpu-ubuntu-16.04
+
+docker build \
+    --build-arg BASE_IMAGE="${BASE_IMAGE}" \
+    --build-arg TENSORFLOW_VERSION="${TENSORFLOW_VERSION}" \
+    --tag "${TAG}" \
+    .
+```
+
+#### GPU example
+
+The following snippet builds a GPU-enabled image on Ubuntu 16.04 with CUDA 9.0:
+
+```shell
+BASE_IMAGE=nvidia/cuda:9.0-cudnn7-runtime-ubuntu16.04
+TENSORFLOW_VERSION=tensorflow-gpu==1.12.0
+TAG=voxel51/eta:gpu-ubuntu-16.04-cuda-9.0
+
+docker build \
+    --build-arg BASE_IMAGE="${BASE_IMAGE}" \
+    --build-arg TENSORFLOW_VERSION="${TENSORFLOW_VERSION}" \
+    --tag "${TAG}" \
+    .
+```
+
+
+## Using an image
+
+After you have built an ETA image, you run the image with an interactive
+shell as follows:
+
+```shell
+# For CPU images
+docker run -it $TAG
+
+# For GPU images
+docker run -it --runtime=nvidia $TAG
+```
+
+
+## Copyright
+
+Copyright 2017-2020, Voxel51, Inc.<br>
+voxel51.com
+
+Brian Moore, [email protected]