PoseClassifier-yolo
Classify human poses with help of yolo pose model.
🖼️ Sample Output
Let's Get Started...
Using this Custom Pose Classification, I am going to Create a Yoga Pose Classification. Using Yoga Poses Dataset.
Clone this Repository
git clone https://github.com/naseemap47/PoseClassifier-yolo.git
cd PoseClassifier-yolo
Install Dependency
Recommended:
conda create -n pose python=3.9 -y
conda activate pose
conda install pytorch==2.0.0 torchvision==0.15.0 torchaudio==2.0.0 pytorch-cuda=11.8 -c pytorch -c nvidia -y
pip3 install -r requirements.txt
OR
pip3 install -r requirements.txt
🎒 Prepare Dataset
Dataset Structure:
├── Dataset
│ ├── class1
│ │ ├── 1.jpg
│ │ ├── 2.jpg
│ │ ├── ...
│ ├── class2
│ │ ├── 1.jpg
│ │ ├── 2.jpg
│ │ ├── ...
. .
. .
🤖 Train
🗃️ Create Landmark Dataset for each Classes
Convert pose images into pose lankmark and save to an CSV file.
So that we can train with that.
Args
-p, --pose: choose yolov8 pose model
Choices:
yolov8n-pose, yolov8s-pose, yolov8m-pose, yolov8l-pose, yolov8x-pose, yolov8x-pose-p6
-i, --data: path to data Dir
-o, --save: path to save csv file, eg: dir/data.csv
Example:
python3 generate_csv.py --pose yolov8n-pose --data dataset/train_data --save data.csv
🚆 Create DeepLearinng Model to predict Human Pose
Create a keras model to predict human poses.
Args
-i, --data: path to data Dir
Example:
python3 train.py --data data.csv
📺 Inference
Inference your Pose model.
Support
- Image
- Video
- Camera
- RTSP
Args
-p, --pose: choose yolov8 pose model
Choices:
yolov8n-pose, yolov8s-pose, yolov8m-pose, yolov8l-pose, yolov8x-pose, yolov8x-pose-p6
-m, --model: path to saved keras model
-s, --source: video path/cam-id/RTSP
-c, --conf: model prediction confidence (0<conf<1)
--save: to save video
--hide: hide video window
Example:
python3 inference.py --pose yolov8n-pose --weight /runs/train4/ckpt_best.pth --source /test/video.mp4 --conf 0.66 # video
--source /test/sample.jpg --conf 0.5 --save # Image save
--source /test/video.mp4 --conf 0.75 --hide # to save and hide video window
--source 0 --conf 0.45 # Camera
--source 'rtsp://link' --conf 0.25 --save # save RTSP video stream