To use this demonstration, the TensorFlow Object Detection API should be used. For more details here.
To use this example you will need these minimum elements of TensorFlow Object Detection API:
Object model, quick option to automatize the deploy:
git clone https://github.com/tensorflow/models.git
*validate dependency with protoc tool
protoc ./models/research/object_detection/protos/string_int_label_map.proto --python_out=.
cp -R models/research/object_detection/ object_detection/
rm -rf model
You can use the preferred model: faster_rcnn_inception_resnet_v2_atrous_coco_2017_11_08 or as another like faster_rcnn_inception_v2_coco_2017_11_08
Note: You should have the Google Cloud SDK. More information about App Engine Flexible environment, Python here
Local: python main.py
Production Environment: gcloud app deploy
*(-v version) if you want to deploy it to a specific version.