Preliminary screening of knee osteoarthritis

Setup

_Instructions refer to Windows-based systems

Please download this compressed package before using https://drive.google.com/drive/folders/1nATFeytb5y11dNfpT_6nRA3pdo8sv6ux?usp=sharing .

Please unzip yolov3-spp.weights to joints_detectors\Alphapose\models\yolo.

Please decompress duc_se.pth to joints_detectors\Alphapose\models\sppe.

Please extract Dataset.rar to Dataset Please put pretrained_h36m_detectron_coco.bin, PSTMO_no_refine_6_4215_h36m_cpn.pth, PSTMO_no_refine_11_4288_h36m_cpn.pth, PSTMO_refine_6_4215_h36m_cpn.pth, PSTMOS_no_refine_15_2936_h36m_gt.pth, PSTMOS_no_refine_28_4306_h36m_cpn.pth, PSTMOS_no_refine_48_5137_in_the_wild.pth, PSTMOS_no_refine_50_3203_3dhp.pth are decompressed under checkpoint

Example

If you want to extract 3D bone information from your video, you can refer to this example. Please open run.py and change VIDEO_path to the local path of your video. Open npload.py, savepath will be the storage location of your human key point time series data, you can modify it according to your actual situation. Open angel.py, savepath will be the storage location of your angle time series data, you can modify it according to your actual situation, you can also increase or decrease the code to select any joint angle time series data you want. Run run.py, you can find your result data in the above path in turn, the data including all key points of the human body is stored under /outputs by default, you can modify the path in videopose_PSTMO.py according to your actual situation to change the save path.

Run the model

Please make sure you extract Dataset.rar to /Dataset. Open any .ipynb file you want to run, confirm that there is no problem with the path of the dataset, and run it to get the result.

Qualitative analysis and quantitative analysis

First locate the DWT analysis directory, download the compressed package from Google Cloud Disk and extract it to the directory, and run the .ipynb file to obtain the analysis report.

https://drive.google.com/file/d/1xosX7hiPJllVZb8KQPXVCXbUYWxQabyl/view?usp=sharing

Remark

we use python3.9,torch1.9.1,torchvision0.10.1. For the model, we used tensorflow2.12.0.