wqael/tf_objdet repository overview

⁠Object Detection API of TensorFlow

This document describes procedures to install and run [1] with a companion docker image. The companion docker image contains all the software dependencies. For runtime, repo of [1] sits on the host system and is mounted into the docker image.

[1] https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/installation.md⁠

⁠Installation

1. Pull the docker image. Dockerfile⁠.

docker pull wqael/tf_objdet:gpu

1. Download the tensorflow/model project: https://github.com/tensorflow/models⁠
2. Mount the models folder into the docker image.

docker run -it -p 8888:8888 -p 6006:6006 -v ~/models:/notebooks wqael/tf_objdet:gpu bash

1. From here on, execute from inside docker's bash, e.g.,

# cd /notebooks/
# ls
AUTHORS  CODEOWNERS  CONTRIBUTING.md  ISSUE_TEMPLATE.md  LICENSE  README.md  WORKSPACE  official  research  samples  tutorials

1. Follow the COCO API installation in [1]
2. Follow the Protobuf compilation in [1]
3. Add Libraries to PYTHONPATH as [1]
4. "Testing the Installation" as [1]

⁠Run-time

1. Mount the models folder into the docker image.

docker run -it -p 8888:8888 -p 6006:6006 -v ~/models:/notebooks wqael/tf_objdet:gpu bash

1. From here on, execute from inside docker's bash, e.g.,

cd /notebooks/research
# From tensorflow/models/research/
export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/slim

1. (Optional) Launch jupyter

cd /notebooks
jupyter notebook --allow-root