An open and extensible computer vision, machine learning and video analytics infrastructure.

This project is supported by the NIST Public Safety Innovation Accelerator Program.

The ETA package requires the following external dependencies:

• OpenCV
• TensorFlow
• ffmpeg
• ImageMagick

ETA is very portable:

• Installable on Mac or Linux
• Supports both Python 2.7.X and Python 3.6.X
• Supports both OpenCV 2.4+ and OpenCV 3.0+
• Supports both CPU and GPU execution
• Supports both CUDA 8 and CUDA 9 (for GPU installations)

0. Activate a virtual environment for this installation (optional)

1. Clone the repository:

git clone https://github.com/voxel51/eta
cd eta

2. Install external dependenices:

bash install_externals.bash

Depending on your Python environment, you may need to run the script with sudo privileges. Note that the install script supports flags that control things like (on macOS) whether port or brew is used to install packages. Run bash install_externals.bash -h for more information.

The script inspects your system to see if CUDA is installed, and, if it is, TensorFlow is installed with GPU support. In particular, if CUDA 9 is found, the latest version of the tensorflow-gpu package is installed, and if CUDA 8 is found, tensorflow-gpu 1.4 is installed.

3. Install the ETA package:

# Install in global (non-editable) mode
pip install .
cd ..
rm -rf eta

# Install in development mode
pip install -e .

When the root eta package is imported, it tries to read the eta/config.json 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, copy the example configuration file

cp config-example.json config.json

and then edit your config file to provide the paths to the relevant directories in your installation. If you do not require a customized installation, the example configuration file contains the pattern {{eta}} that you can perform a quick find-and-replace on to populate the config:

# on a mac
sed -i '' -e "s|{{eta}}|$(pwd)|g" config.json

# on most linux distributions
sed -i -e "s|{{eta}}|$(pwd)|g" config.json

The default config includes the eta/modules, eta/pipelines, and eta/models directories on your module, pipeline, and models search paths, respectively. These directories contain the necessary information to run the standard analytics exposed by the ETA library. In addition, the relative paths ./modules, ./pipelines, and ./models are added to their respective paths to support the typical directory structure that we adopt for our custom projects.

When you installed ETA, an eta command-line utility was added to your path. This utility provides access to many usefuel features of ETA, including building pipelines from requests, running pipelines, and generating module metadata files.

To learn more about the supported operations and their syntaxes, type

eta --help

To test your installation, run the following commands:

# Download example data
python examples/download_data.py

# Perform some image manipulation
python examples/demo_images.py

# Run a pre-configued video processing pipeline
eta run examples/demo_video/pipeline.json

# Build and run pipelines from requests using the `eta` command-line tool
eta build -r examples/demo_video_formatter/request.json --run-now
eta build -r examples/demo_video_clipper/request.json --run-now

#
# Example image/video embedding
#
# NOTE: A 550MB VGG16 weights file will be downloaded from the web and stored
#       in eta/models/ the first time you run one of these!
#
cd examples/demo_embed_vgg16
# Example image embedding
python embed_image.py
# Another example image embedding
python embed_image_direct.py
# Example video embedding
python embed_video.py
# Example use of the embed_vgg16 module
bash embed_vgg16_module.bash
# Example embedding pipeline
eta run embed_vgg16_pipeline-config.json

You can use virtual environments to maintain a separate Python working environment for ETA that operates independently of other packages and Python applications on your machine. See docs/virtualenv_guide.md for more details and setup instructions.

pip uninstall eta

Copyright 2017-2019, Voxel51, Inc.
voxel51.com

Brian Moore, brian@voxel51.com