Source from 요룰레히 9_21- HEYYEYAAEYAAAEYAEYAA (What's Up)
This is a simply application for using OpenPose model for detecting hands and showing the heatmap.
You can get the pre-trained model from Hugging Face.
Download it into ./model
.
https://huggingface.co/NatLee/openpose-keras-model
git clone --depth=1 [email protected]:NatLee/openpose-keras-model ./model/
model
└── weights.h5
Tested on x86/64 CPU, Intel 12th-i5 with 16 GB RAM.
-
Prepare the model from #Model.
-
Prepare your video in
./data
namedtest.mp4
. The directory will like below:. ├── data │ ├── README.md │ └── test.mp4 <--- Your video ├── doc │ ├── heatmap-motion.gif │ ├── heatmap.png │ └── original-frame.png ├── docker-compose.yml ├── Dockerfile ├── LICENSE ├── model │ └── weights.h5 <--- The model ├── README.md ├── requirements.txt └── src ├── convert-png-to-gif.py ├── docker-entrypoint.sh ├── main.py ├── openpose │ ├── __init__.py │ ├── model.py │ ├── OpenposeModel.py │ ├── prediction.py │ └── util.py └── utils ├── draw_heat.py ├── FPS.py ├── heatmapper │ ├── assets │ │ ├── 450pxdot.png │ │ ├── default.png │ │ └── reveal.png │ ├── heatmapper.py │ └── __init__.py ├── __init__.py └── ROIs.py 8 directories, 28 files
-
Just run docker command with the following:
docker-compose build && docker-compose up
The result will be generated in the folder
./tmp
.
Nat Lee |
- Heatmap package is from HeatMappy.
- The version of keras with Open Pose model is from Open-Pose-Keras.