Humanposture Classification

human-pose-series-classification

• 3 Types of models of human-pose-series-classification.

Install

• Address: https://pan.baidu.com/s/1WACOLaTOQqR3s-DEjAgdCw Password: vzxh
• CNN data: data_for_CNN_and_Transformer.rar
• CNN model: CNNmodel.pth
• RNN(lstm) data: RNN_data.zip
• RNN(lstm) model: model_1000EPOch_0.77test.pkl
• Transformer data: data_for_CNN_and_Transformer.rar
• Transformer model: ./checkpoint/model.pkl

Usage

CNN train:

• CNN3.py is the training code

CNN test:

• CNNfinals.py uses it directly to test

RNN train:

• Using the data in the root of the code, you can use the jupyter notebook to step up and implement
• (and you can see the previous training records in jupyter directly, and I've trained to test accuracy= 0.8).

RNN test:

• In the last three modules of RNN(LSTM)Train_Test.ipynb, we can directly load the model model_1000EPOch_0.77test.pkl to test the results by making sure that the model of LSTM is defined and the test data set and its initialization (initial and detach) are loaded

Transformer train:

• Place your dataset in the file data, split them into train/test data and place them in the subfolder train and test.
• Run main.py, then the model will be saved in folder checkpoints as model.pkl.

Transformer test:

• Place the pretrained model in folder checkpoints and name it as model.pkl.

• Open main.py and comment line 178

  path = train(device, train_loader=train_loader, valid_loader=valid_loader, epochs=epochs)

• Run main.py, then it will show test accuracy with your test data.