DeepOrchestration

It is a cool project using deep learning to help ochestration. Fighting!

The online editing link is here:

https://www.overleaf.com/4125771964jsvptcvnzdwj

Generate the dataset

• download the raw data from http://www.orch-idea.org/ (I choose StaticSOL but you can choose other datasets)
• select all ord files played by ['Vc', 'Fl', 'Va', 'Vn', 'Ob', 'BTb', 'Cb', 'ClBb', 'Hn', 'TpC', 'Bn', 'Tbn'] and put these files in a shared directory
• call method show_all_class_num in process_OrchDB.py to get the class.index
• set paras in process_OrchDB.py
• call method random_combine to generate new datasets

Train a network

Just run python main.py. Before training, make sure that the class number and the dictionary db are right. If it is a new dataset, call method show_all_class_num in process_OrchDB.py to get the new class number and call method stat_test_db in process_OrchDB.py to get the new db.