Make sure clone all the file inside this github page.
Some importance data、model、video put inside the GoogleDrive.
Put all the files inside the same path of the directories.
Devices:
- Ubuntu 16.04 LTS
- Memory: 15.6 GiB
- Processor: Intel® Core™ i5-9400F CPU @ 2.90GHz × 6
- Graphics : GeForce RTX 2080/PCIe/SSE2
- Os Type: 64-bit
Dependency:
- Cuda compilation tools, release 10.0, V10.0.130
- Cudnn : 7.4.2
- Python : 3.6.12
- tensorflow-gpu: 1.15.0
- keras : 2.3.1
Check Cuda Version:
nvcc -V
Check Cuda Version:
cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2
or
cat /usr/include/cudnn.h | grep CUDNN_MAJOR -A 2
Make sure already install Anaconda.
if not:
- Download the { }.sh file. here
- Install Anaconda:
bash ./{filename.sh}
Completed install Anacconda:
- Create new environment and activate:
conda create -n {env_name} python=3.6.12
conda activate {env_name}
- Install tensorflow-gpu 1.15.0 version(make sure your tensorflow-gpu version is fit to your cuda version)
pip3 install tensorflow-gpu==1.15.0
- Check if your tensorflow-gpu completed install or not.
python
import tensorflow as tf
tf.__version__
1.15.0
cd src/depends/tf-pose-estimation
pip3 install -r requirements.txt
MobileNet-Thin models has been inside, see folder:
src/depends/tf-pose-estimation/models/graph
cmu mobilenet_thin mobilenet_v2_large mobilenet_v2_small
Choose to download(CMU models)
Original Version more accuracy but large and long execution time:
cd models/graph/cmu
bash download.sh
cd ../../../
pip3 install jupyter tqdm
pip3 install keras==2.3.1
sudo apt install swig
pip3 install "git+https://github.com/philferriere/cocoapi.git#egg=pycocotools&subdirectory=PythonAPI"
cd src/depends/tf-pose-estimation/tf_pose/pafprocess
swig -python -c++ pafprocess.i && python3 setup.py build_ext --inplace
cd ../../../../../
pip3 install -r requirements.txt
Test on video:
python src/s5_test.py \
--model_path model/trained_classifier.pickle \
--data_type video \
--data_path data_test/exercise.avi \
--output_folder output
if error occurred!
cd src/depends/tf-pose-estimation
python setup.py install
Then try Final check again.
- Chenge Yang, Zhicheng Yu, Feiyu Chen, "Human Pose Estimation Benchmarking and Action Recognition," Deep Learning Project, Winter 2019, Northwestern University, 26-Oct-2019.
- Feiyu Chen, "Human-Pose-Estimation-Benchmarking-and-Action-Recognition,"Deep Learning Project, Winter 2019, Northwestern University, 26-Oct-2019.
- jiajunhua, "ildoonet-tf-pose-estimation," 10-Apr-2019
- philferriere ,"cocoapi,"25-Oct-2018
- 橘白卯咪,"欸你在幹嘛?跟一眼瞬間就懂你的行為辨識技術初次見面," 01-Sep-2020
- Zhe Cao, Gines Hidalgo, Tomas Simon, Shih-En Wei, Yaser Sheikh,"OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields,"Computer Vision and Pattern Recognition,18-Dec-2018