Added data parameters script

Author: Svito-zar

data_processing/data_params.py

@@ -0,0 +1,33 @@
+"""
+Shared argument parser for data_params.py and create_vector.py. 
+
+By default, we assume that the dataset is found in the <repo>/dataset/raw/ folder,
+and the preprocessed datasets will be created in the <repo>/dataset/processed/ folder,
+but these paths can be changed with the parameters below.
+"""
+import argparse
+
+parser = argparse.ArgumentParser(
+    description="""Parameters for data processing for the paper `Gesticulator: 
+                   A framework for semantically-aware speech-driven gesture generation""")
+
+# Folders params
+parser.add_argument('--raw_data_dir', '-data_raw', default="../../dataset/raw/",
+                    help='Path to the folder with the raw dataset')
+parser.add_argument('--proc_data_dir', '-data_proc', default="../../dataset/processed/",
+                    help='Path to the folder with the processed dataset')
+
+# Sequence processing
+parser.add_argument('--seq_len', '-seq_l', default=40,
+                    help='Length of the sequences during training (used only to avoid vanishing gradients)')
+parser.add_argument('--n_context', '-n_cont', default=60, type=int,
+                    help='Length of a total past and future context for speech to be used to generate gestures')
+
+# Features
+parser.add_argument('--feature_type', '-feat', default="MFCC",
+                    help='''Describes the type of the input features 
+                            (can be \'Spectro\', \'MFCC\', \'Pros\', \'MFCC+Pros\' or \'Spectro+Pos\')''')
+parser.add_argument('--n_input', '-n_in', default=26, type=int,
+                    help='Number of inputs to the model')
+parser.add_argument('--n_output', '-n_out', default=45, type=int,
+                    help='number of outputs of the model')