This repository is meant to be used as a starting point for object detection tasks involving the TensorFlow object detection API.
Before to be able to use the TensorFlow object detection API, they are several steps that needs to be followed.
To installed an already prepared conda environment, run:
conda create --name myenv --file spec-file.txt
Clone this GitHub repository somewhere in your home folder.
You also need to update your PYTHONPATH variable to include the TensorFlow object detection API import.
export PYTHONPATH=$PYTHONPATH:/folder_where_you_cloned_the_above_repo/models/research:/folder_where_you_cloned_the_above_repo/models/research/slim
You will eventually need to run it every time you open a new terminal.
Finally the TensorFlow object detection API framework needs to be compiled. Go into the models/research
folder and run :
protoc object_detection/protos/*.proto --python_out=.
To test the framework is correcly installed simply run from the models/research folder:
python object_detection/builders/model_builder_test.py