CES_intent_demo

Version info

Python : 2.7.12, CUDA : 8.x, CuDNN : 5.1.10

Installation

Download C3D and I3D codes from github repositories, locate 'i3d.py' under the '/origin' and 'optical_flow.py'(opencv samples, https://github.com/opencv/opencv) under the '/origin' and 'intent_tf_local', and set the paths in the codes. Download darknet codes from github repository, locate under the '/src/vintent'.

git clone https://github.com/hx173149/C3D-tensorflow.git
git clone https://github.com/deepmind/kinetics-i3d.git
git clone https://github.com/AlexeyAB/darknet

TRAIN the model

cd ./orgin
python train_intent_model.py

STEPS TO RUN THIS PROGRAM

# step 1 : run computation server (grpc)
cd ./src
python server.py

# step 2 : run flask app for web demo (just for html rendering)
cd ./intent_tf_local
python flask_app.py

# step 3 : run client.py('0' means cam index)
python client.py 0

# step 4 : visit 'localhost:5001' with Chrome
...

• If you want to build src/* as a new docker, deploy/deploy.sh will do the whole process of 'build image->create & run container'

cd deploy
bash deploy.sh

• If you want to start already-existing container 'vintent_container',

nvidia-docker start vintent_container