Data and notebooks to print skeletons added

notebooks/user_data_loader.py

@@ -0,0 +1,147 @@
+import numpy as np
+import pandas as pd
+import arff as arff   # Downloaded from: http://code.google.com/p/arff/
+# import matplotlib.pyplot as plt
+import itertools as it
+import toolz as tlz
+from functools import partial
+
+# tuple with the names of each attribute
+# The last one ('pose') defines the class
+# index = ('h_seqNum', 'h_stamp', 'user_id', 
+#          'head_confidence', 'head_pos_x', 'head_pos_y', 'head_pos_z', 
+#          'head_orient_x', 'head_orient_y', 'head_orient_z', 'head_orient_w', 
+#          'neck_confidence', 'neck_pos_x', 'neck_pos_y', 'neck_pos_z', 
+#          'neck_orient_x', 'neck_orient_y', 'neck_orient_z', 'neck_orient_w', 
+#          'torso_confidence', 'torso_pos_x', 'torso_pos_y', 'torso_pos_z', 
+#          'torso_orient_x', 'torso_orient_y', 'torso_orient_z', 'torso_orient_w', 
+#          'left_shoulder_confidence', 'left_shoulder_pos_x', 'left_shoulder_pos_y', 'left_shoulder_pos_z', 
+#          'left_shoulder_orient_x', 'left_shoulder_orient_y', 'left_shoulder_orient_z', 'left_shoulder_orient_w', 
+#          'left_elbow_confidence', 'left_elbow_pos_x', 'left_elbow_pos_y', 'left_elbow_pos_z', 
+#          'left_elbow_orient_x', 'left_elbow_orient_y', 'left_elbow_orient_z', 'left_elbow_orient_w', 
+#          'left_hand_confidence', 'left_hand_pos_x', 'left_hand_pos_y', 'left_hand_pos_z', 
+#          'left_hand_orient_x', 'left_hand_orient_y', 'left_hand_orient_z', 'left_hand_orient_w', 
+#          'right_shoulder_confidence', 'right_shoulder_pos_x', 'right_shoulder_pos_y', 'right_shoulder_pos_z', 
+#          'right_shoulder_orient_x', 'right_shoulder_orient_y', 'right_shoulder_orient_z', 'right_shoulder_orient_w', 
+#          'right_elbow_confidence', 'right_elbow_pos_x', 'right_elbow_pos_y', 'right_elbow_pos_z', 
+#          'right_elbow_orient_x', 'right_elbow_orient_y', 'right_elbow_orient_z', 'right_elbow_orient_w', 
+#          'right_hand_confidence', 'right_hand_pos_x', 'right_hand_pos_y', 'right_hand_pos_z', 
+#          'right_hand_orient_x', 'right_hand_orient_y', 'right_hand_orient_z', 'right_hand_orient_w', 
+#          'left_hip_confidence', 'left_hip_pos_x', 'left_hip_pos_y', 'left_hip_pos_z', 
+#          'left_hip_orient_x', 'left_hip_orient_y', 'left_hip_orient_z', 'left_hip_orient_w', 
+#          'left_knee_confidence', 'left_knee_pos_x', 'left_knee_pos_y', 'left_knee_pos_z', 
+#          'left_knee_orient_x', 'left_knee_orient_y', 'left_knee_orient_z', 'left_knee_orient_w', 
+#          'left_foot_confidence', 'left_foot_pos_x', 'left_foot_pos_y', 'left_foot_pos_z', 
+#          'left_foot_orient_x', 'left_foot_orient_y', 'left_foot_orient_z', 'left_foot_orient_w', 
+#          'right_hip_confidence', 'right_hip_pos_x', 'right_hip_pos_y', 'right_hip_pos_z', 
+#          'right_hip_orient_x', 'right_hip_orient_y', 'right_hip_orient_z', 'right_hip_orient_w', 
+#          'right_knee_confidence', 'right_knee_pos_x', 'right_knee_pos_y', 'right_knee_pos_z', 
+#          'right_knee_orient_x', 'right_knee_orient_y', 'right_knee_orient_z', 'right_knee_orient_w', 
+#          'right_foot_confidence', 'right_foot_pos_x', 'right_foot_pos_y', 'right_foot_pos_z', 
+#          'right_foot_orient_x', 'right_foot_orient_y', 'right_foot_orient_z', 'right_foot_orient_w', 
+#          'pose')
+
+users = ['user' + str(i+1).zfill(2) for i in np.arange(30)]
+users.pop(users.index('user12')) # There is no 'user12' data, so we remove it
+
+header = tuple(['h_seqNum', 'h_stamp', 'user_id'])
+joints = tuple(['head', 'neck', 'torso', 
+                'left_shoulder', 'left_elbow', 'left_hand', 
+                'right_shoulder', 'right_elbow', 'right_hand', 
+                'left_hip', 'left_knee', 'left_foot', 
+                'right_hip', 'right_knee', 'right_foot'])
+attribs = tuple(['confidence', 'pos_x','pos_y','pos_z',
+                 'orient_x','orient_y','orient_z','orient_w'])
+
+index = list(it.chain(header, 
+                      it.imap('_'.join, it.product(joints, attribs)), 
+                      ['pose',]))
+
+positions = [i for i in index if '_pos_' in i]
+orientations = [i for i in index if '_orient_' in i]
+confidences = [i for i in index if '_confidence' in i]
+pose = index[-1]
+
+ind_pos_x = [i for i in index if '_pos_x' in i]
+ind_pos_y = [i for i in index if '_pos_y' in i]
+ind_pos_z = [i for i in index if '_pos_z' in i]
+
+def make_multiindex(joints, attribs):
+    '''Returns all attribs in two iterables.
+       1st iterable is the 1st level of the index. 
+       2nd iterable is the 2nd level of the index.'''
+    all_columns = it.izip(*it.product(joints, attribs))
+    first_level = it.chain(['header']*len(header),all_columns.next(), ['pose',])
+    second_level = it.chain(header, all_columns.next(), ['pose',])
+    return first_level, second_level
+
+
+def normalize_joints(df, from_joint):
+    ''' Returns a normalized DataFrame. '''
+#    df_orig = df[from_joint].copy()
+#    df_orig.columns = pd.MultiIndex.from_arrays([[from_joint+'_orig', ]*len(attribs), attribs])
+    return df[from_joint].copy(), df.drop(['header', 'pose', from_joint], axis=1, level='joint').sub(df[from_joint], level=1)    
+
+def load_user_file(file_):
+    ''' Loads an ARFF file from an experiment and returns a DataFrame associated to it '''
+    data = arff.load(file_)
+    # Converting it to a Numpy array
+    data = np.array([list(d) for d in data])
+    # Converting the array to a pandas dataFrame
+    df = pd.DataFrame(data, columns=index) 
+    # Setting dtype to float for the numeric columns
+    numeric_cols = list(index[3:-1])
+    df[numeric_cols] = df[numeric_cols].astype(float)
+    return df
+
+# Function to clean the index removing the leading 'STAND_' substring 
+stand_cleaner = lambda x: x.lstrip('STAND_') 
+
+def prepare_df(df):
+    ''' Groups by pose, removes Header and Confidences, 
+        calculates the mean for all instances of each group, 
+        and cleans the index names '''
+    return df.groupby(pose).mean().drop(confidences, axis=1).rename(index=stand_cleaner)
+
+def accumulate_indices(df_list):
+    ''' Concatenates the indices of a list of pandas.Dataframes and returns them as an numpy.array '''
+    get_ind = lambda x: x.index
+    return np.concatenate(map(get_ind, df_list))
+
+def accumulate_users(users):
+    ''' Accumulates a list of dataframes in a single dataframe '''
+    return pd.concat(users) 
+
+def insert_user_name(df, user):
+    ''' returns a copy of the entered dataframe, but with a column added whith the user id '''
+    new_df = df.copy()
+    new_df.insert(0 , 'user', [user] * len(df))
+    return new_df
+
+# parse_filename = lambda user, experiment, dir_: \
+#                         dir_ + '/' + experiment + '-' + user + '.arff'
+def parse_filename(user, experiment, dir_):
+    return dir_ + '/' + experiment + '-' + user + '.arff'
+
+parse_exp03_filename = partial(parse_filename, experiment='exp03', 
+                               dir_='../data/users_separated',)
+
+# user_pipe = lambda filename: tlz.pipe(filename, load_user_file, prepare_df)
+# user_pipe.__doc__= ''' A pipe that produces a user dataset (pd.DataFrame) 
+#                         starting from a filename. 
+#                        Note that the data in the dataset corresponds 
+#                        to the means of data found in the file '''
+def user_pipe(filename):
+    ''' A pipe that produces a user dataset (pd.DataFrame) 
+        starting from a filename. 
+        Note that the data in the dataset corresponds 
+        to the means of data found in the file '''
+    return tlz.pipe(filename, load_user_file, prepare_df)
+
+
+def load_all_users():
+    ''' Returns a pd.DataFrame with the information of all the users'''
+    map = tlz.curry(map)
+    dataset = tlz.pipe(users, map(parse_exp03_filename), map(user_pipe), accumulate_users)
+    dataset.insert(0, 'user', sorted(users * 3))
+    return dataset