A Comparative Analysis of IMU based Motion Prediction Methods

To explore the effects of the different types of network architectures on the human pose estimation, we implement four different architectures: (1) Bidirectional LSTM, (2) Unidirectional LSTM, (3) Bidirectional GRU, (4) Unidirectional GRU. We train models based on these architectures using the DIP-IMU Dataset.

The DIP-IMU dataset can be downloaded from the following link:

Link to DIP_IMU dataset

Unzip the dataset in the folder ./AML_project

The Bidirectional LSTM model can be trained as follows:

python3 DIP_IMU_NN_BiRNN.py --run-name lstm_bi --network LSTM --bidirectional --train --epochs 30

For training the other models, you can use the arguments we provided in ./hyperparameters.yaml.

You can change the number of epochs for which the model is trained to observe the change in performance.

If you just wish to plot the data obtained from previously trained models, use the following command:

python3 DIP_IMU_NN_BiRNN.py --run-name <name_of_run_you_wish_to_plot>