saliency/loaddata.py at main · crispchris/saliency · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
"""
load and view dataset
For Dataset UCR and Dataset Tool Tracking

"""

## -------------------
## --- Third-Party ---
## -------------------
import sys
sys.path.append('..')
import argparse
import numpy as np
import matplotlib.pyplot as plt

## ---------------------
## --- Tool-Tracking ---
## ---------------------
from datatools import ACC, GYR, MAG, MIC, POS, VEL

## -----------
## --- Own ---
## -----------
from utils import plot_dataset, read_dataset_ts, plot_sample_per_dimension, plot_samples_each_class
from utils import create_directory
from tool_tracking.dataknowing.loadData import load_data, pipe_segment, filteroutlabels, balance_classes_oneinput
from tool_tracking.dataknowing.loadData import summarize_labels, balance_classes, split_data, split_data_oneinput
from tool_tracking.dataknowing.loadData import merge_labels, filteroutlabels_one_array, zero_normalization
from tool_tracking.tool_utils import Traindata


# if __name__ == '__main__':
## Arguments
parser = argparse.ArgumentParser(description="Datasets View")
parser.add_argument("--dataset", type=str, default="tool_tracking", help="the name of Dataset")
parser.add_argument("--tool", type=str, default="electric_screwdriver", help="the tool name if tool_tracking is used")
parser.add_argument("--window", type=bool, default=True, help="whether window the samples if tool_tracking or stabilo is used")
parser.add_argument("--window_len", type=float, default=0.2, help="window length if window is used in tool_tracking or stabilo is used")
parser.add_argument("--window_overlap", type=float, default=0.5, help="window overlap if window is used in tool_tracking or stabilo is used")
parser.add_argument("--plot_dimension", type=bool, default=False, help="Plot classes per dimension")
parser.add_argument("--plot_one_sample", type=bool, default=True, help="Plot a sample per class")
parser.add_argument("--plot_samples", type=bool, default=False, help="Plot some samples in one plot per class")

args = parser.parse_args()
dataset_name = args.dataset
dataset_name = "GunPointAgeSpan"
# Load the dataset
if dataset_name is "tool_tracking":
    ## here only 4 classes will be viewed
    ## classes = {2, 3, 4, 5}
    source = "data/tool-tracking/tool-tracking-data"
    tool = args.tool
    sensors = ['acc', 'gyr', 'mag']
    sensors_list = [ACC, GYR, MAG]
    garbage_labels = [-1, 6, 7, 8, 14]
    label_summary = {0:[2], 1:[3], 2:[4], 3:[5]}
    ## set window length and overlap
    if args.window:
        window_length = args.window_len
        window_overlap = args.window_overlap

    ## Load Data
    data_dict = load_data(source=source, tool=tool, datatypes=sensors_list) ## Each measurement is a DataBunch
    ## Windowing, into one array
    dataset, yt = pipe_segment(data=data_dict, window_length=window_length, overlap=window_overlap,
                               enforce_size=True, sensors=sensors, discard_time=True,
                               one_array=True)

    ## Filter out garbage labels
    dataset, yt = filteroutlabels_one_array(dataset, labels=yt, garbage_labels=garbage_labels)
    reYt = merge_labels(yt, label_mapping=label_summary, print_info=True)

    ## zero mean normalization
    data = None
    for i in range(dataset.shape[0]):
        if data is None:
            data = dataset[i]
        else:
            data = np.concatenate((data, dataset[i]), axis=-1)

    std_ = np.std(data, axis=-1).reshape((1, -1, 1))
    mean_ = np.mean(data, axis=-1).reshape((1, -1, 1))
    norm_dataset = (dataset - mean_) / std_

    ## for plots
    labels = ["ACC_X-axis", "ACC_Y-axis", "ACC_Z-axis",
              "GYR_X-axis", "GYR_Y-axis", "GYR_Z-axis",
              "MAG_X-axis", "MAG_Y-axis", "MAG_Z-axis"]
    if args.plot_dimension or args.plot_one_sample:
        root_dir = "results/" + dataset_name + "/dataview/"
        create_directory(root_dir)

        samples = []
        y = []
        for key in label_summary:
            np.random.seed(key)
            mask = reYt == key
            rand = np.random.randint(0, len(reYt[mask]), size=1)
            y.append(reYt[mask][rand])
            samples.append(norm_dataset[mask][rand][0])

        if args.plot_dimension:
        # for i in range(len(y)):
            for j in range(samples[0].shape[0]):
                plt.figure()
                plt.plot(range(samples[0].shape[-1]), samples[0][j, :], label=f"class {y[0][0]}")
                plt.plot(range(samples[1].shape[-1]), samples[1][j, :], label=f"class {y[1][0]}")
                plt.plot(range(samples[2].shape[-1]), samples[2][j, :], label=f"class {y[2][0]}")
                plt.plot(range(samples[3].shape[-1]), samples[3][j, :], label=f"class {y[3][0]}")
                plt.legend(loc='best')
                plt.title(f"{labels[j]} sensor for {len(y)} classes")
                file_name = root_dir + f"{labels[j]}_sensor.png"
                plt.savefig(file_name)
            plt.show()
        if args.plot_one_sample:
            for i in range(len(y)):
                plt.figure()
                plt.plot(range(samples[i].shape[-1]), samples[i][0, :], label=f"Sensor {labels[0]}")
                plt.plot(range(samples[i].shape[-1]), samples[i][1, :], label=f"Sensor {labels[1]}")
                plt.plot(range(samples[i].shape[-1]), samples[i][2, :], label=f"Sensor {labels[2]}")
                plt.plot(range(samples[i].shape[-1]), samples[i][3, :], label=f"Sensor {labels[3]}")
                plt.plot(range(samples[i].shape[-1]), samples[i][4, :], label=f"Sensor {labels[4]}")
                plt.plot(range(samples[i].shape[-1]), samples[i][5, :], label=f"Sensor {labels[5]}")
                plt.plot(range(samples[i].shape[-1]), samples[i][6, :], label=f"Sensor {labels[6]}")
                plt.plot(range(samples[i].shape[-1]), samples[i][7, :], label=f"Sensor {labels[7]}")
                plt.plot(range(samples[i].shape[-1]), samples[i][8, :], label=f"Sensor {labels[8]}")
                plt.legend(loc='best')
                plt.title(f"a sample for {y[i][0]} class")
                file_name = root_dir + f"a_sample_{y[i][0]}_class.png"
                plt.savefig(file_name)
            plt.show()

    if args.plot_samples:
        root_dir = "results/" + dataset_name + "/dataview/"
        create_directory(root_dir)
        y = []
        samples = []
        for key in label_summary:
            np.random.seed(key)
            mask = reYt == key
            rand = np.random.randint(0, len(reYt[mask]), size=10)
            y.append(reYt[mask][rand])
            samples.append(norm_dataset[mask][rand])

        for i in range(len(y)):
            plt.figure()
            for j in range(len(y[i])):
                plt.plot(range(samples[i][j].shape[-1]), samples[i][j][0, :], label=f"Sensor {labels[0]}")
                plt.plot(range(samples[i][j].shape[-1]), samples[i][j][1, :], label=f"Sensor {labels[1]}")
                plt.plot(range(samples[i][j].shape[-1]), samples[i][j][2, :], label=f"Sensor {labels[2]}")
                plt.plot(range(samples[i][j].shape[-1]), samples[i][j][3, :], label=f"Sensor {labels[3]}")
                plt.plot(range(samples[i][j].shape[-1]), samples[i][j][4, :], label=f"Sensor {labels[4]}")
                plt.plot(range(samples[i][j].shape[-1]), samples[i][j][5, :], label=f"Sensor {labels[5]}")
                plt.plot(range(samples[i][j].shape[-1]), samples[i][j][6, :], label=f"Sensor {labels[6]}")
                plt.plot(range(samples[i][j].shape[-1]), samples[i][j][7, :], label=f"Sensor {labels[7]}")
                plt.plot(range(samples[i][j].shape[-1]), samples[i][j][8, :], label=f"Sensor {labels[8]}")

            plt.legend(loc='upper right', bbox_to_anchor=(0.5, -1.0))
            plt.title(f"samples for {y[i][0]} classes")
            file_name = root_dir + f"{y[i][0]}_classes.png"
            plt.savefig(file_name)
        plt.show()

elif dataset_name is "stabilo":
    pass
else: ## UCR Datasets
    root_dir = ""
    dataset_name = "GunPointOldVersusYoung"
    multivartiate = False
    classes_dict = {'1': "Young", '2': "Old"}
    dataset = read_dataset_ts(root_dir, dataset_name, znorm=True,
                              multivartiate=multivartiate)
    # create_directory("/results/GunPointAgeSpan/dataview/")
    plot_sample_per_dimension(root_dir, dataset, dataset_name,
                              classes_dict=classes_dict)      #plot a random sample per dimension
    # plot_samples_each_class(root_dir, dataset, dataset_name)        #plot random samples for each class in train and testset
    plot_dataset(root_dir, dataset, dataset_name,
                 classes_dict=classes_dict)   # Plot the whole dataset


# if __name__ == "__main__":