This is the code of MAM (Motor Assessment Model) based on Pytorch.
We provide a small dataset in ./prepared_data to demo the codes.
git clone https://github.com/qiang-Blazer/MAM.git
This will be done in 10 seconds.
We run the codes on the Unbuntu 20.04 operating system.
matplotlib==3.6.0
numpy==1.23.1
pandas==2.0.0
PyYAML==6.0
scikit-learn==1.1.2
torch==1.12.1
tqdm==4.64.1
transformers==4.28.1Install required packages:
pip install -r requirements.txt
python pretrain_clip.py
The pre-trained weights will be in ./checkpoints.
This will be down in 5 minutes for the demo dataset.
python main.py
The results will be in runs.
python inference.py
| Parameter Name | Type | Default Value | Description |
|---|---|---|---|
| -epoch | int | 300 | Number of epochs |
| -batch_size | int | 32 | Mini-batch size in training |
| -cv_id | int | 0 | The cross-validation ID, choose from [0,1,2,3,4] |
| -joint_in_channels | int | 3 | Input channel numbers of joint feature |
| -joint_hidden_channels | int | 64 | Hidden channel numbers of joint feature |
| -time_window | int | 10 | The size of the sliding window in the time dimension |
| -time_step | int | 1 | The step of the sliding window in the time dimension |
| -optimizer | str | "SGD" | Optimizer used in training |
| -lr | float | 0.001 | The starting learning rate |
| -scheduler | str | "CyclicLR" | Scheduler used in training |
| -seed | int | 0 | The random seed |
| -tl_margin | float | 0.4 | The triplet loss margin |
| -with_info | bool | True | If use the info characteristics |