Skip to content

MSc. Dissertation project submitted to School of Computer Science, University of Birmingham. (2023)

Notifications You must be signed in to change notification settings

lastnameis-borah/Deep-Learning-for-Motion-Capture-Sequence-Prediction-in-Quartet-Performances

Repository files navigation

A run through the repository...

MSc. Dissertation project submitted to School of Computer Science, University of Birmingham. (2023)

  1. 'MoCap Dataset' directory holds all the marked MoCap recordings with .mat file extension.

  2. In the directory 'Pre-processing 1', use the .mat files from the previously mentioned directory and outputs four CSV files for all four instruments. These csv files hold the averaged marker values.

  3. Repeat step 2 until all the files from the MoCap Dataset directory are served.

  4. Then make all the output from step 2 go through the reshaper.ipynb file in the 'Pre-processing 2' directory.

  5. This will upsample or downsample the timeseries to 6000 frames. Repeat step 4 for all the CSVs.

  6. Following this, make the file go through the interpolation.ipynb file in the same directory as step 4.

  7. This will interpolate the gaps in the timeseries. Repeat step 6 for all the CSVs.

  8. Resulting CSVs will be input to the LSTM model.

  9. LSTM model is available in the Model_Sequence.ipynb file.

  10. Step 9 will output the predicted quartet sequence.

  11. Model_Tempo.ipynb holds the model to classify the tempo.

  12. Step 11 marks the end of the proposed project.

Author @Anurag Borah
2023

About

MSc. Dissertation project submitted to School of Computer Science, University of Birmingham. (2023)

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published