cs224w/dataset_loader.py at main · rondubi/cs224w · 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
import utils
import csv
import networkx as nx
import matplotlib.pyplot as plt
import pandas as pd

#data_h5 = utils.read_h5("frames_data_small.h5")

#print(len(data_h5))

csv_path = "metavd_v1.csv"

class GraphLoader:
    def __init__(self, verbose=False, visualization=False) -> None:
        #each key is a label and each value is the label's id and also an array of the video ids
        self.labels = {}
        #never used but is a set of all used node_ids
        self.all_node_ids = set()
        #next id that will get assigned
        self.cur_id = 0
        #actual graph
        self.G = nx.Graph()
        #verbose printing
        self.verbose = verbose

        self.visualization = visualization

    #give nodes an id for graph connections
    def get_id(self):
        self.all_node_ids.add(self.cur_id)
        r = self.cur_id
        self.cur_id+=1
        if self.cur_id % 1000:
            if self.verbose:
                print("loaded id", self.cur_id)
        return r

    #iterate all in h5 and create nodes and store set of all nodes of each type
    def create_labels(self, data_h5):
        for key in data_h5.keys():
            l = data_h5[key]['label'].decode('utf-8')
            dataset = data_h5[key]['dataset'].decode('utf-8')
            unique_label = f'{l}-{dataset}'
            #unique_label = l

            id = self.get_id()
            if unique_label in self.labels:
                self.labels[unique_label]['nodes'].append(id)
            else:
                self.labels[unique_label] = {}
                self.labels[unique_label]['nodes'] = [id]
                label_id = self.get_id()
                self.labels[unique_label]['label_id'] = label_id
                self.labels[unique_label]['dataset'] = dataset
                self.G.add_node(
                    label_id,
                    node_type='label',
                    label=unique_label
                )

            """
            print()
            for attribute in data_h5[key]:
                print(attribute)
                print(type(data_h5[key][attribute]))
            print()
            """
            if not self.visualization:
              self.G.add_node(
                  id,
                  node_type='video',
                  dataset=data_h5[key]['dataset'].decode('utf-8'),
                  label=data_h5[key]['label'].decode('utf-8'),
                  split=data_h5[key]['split'].decode('utf-8'),
                  num_frames=int(data_h5[key]['num_frames']),
                  frame_indices=data_h5[key]['frame_indices'].tolist(),
                  frames=data_h5[key]['frames'].tolist(),
                  raw_path=data_h5[key]['raw_path'].decode('utf-8'),
                  embeddings=data_h5[key]['embeddings'].tolist()
              )
            else:
              self.G.add_node(
                  id,
                  node_type='video',
                  dataset=data_h5[key]['dataset'].decode('utf-8'),
                  label=data_h5[key]['label'].decode('utf-8')
              )
            """
            self.G.add_node(
                id,
                node_type='video',
                dataset=data_h5[key]['dataset'].decode('utf-8'),
                label=data_h5[key]['label'].decode('utf-8'),
                split=data_h5[key]['split'].decode('utf-8'),
                num_frames=int(data_h5[key]['num_frames']),
                frame_indices=data_h5[key]['frame_indices'].tolist(),
                frames=data_h5[key]['frames'].tolist(),
                raw_path=data_h5[key]['raw_path'].decode('utf-8'),
                embeddings=data_h5[key]['embeddings'].tolist()
            )
            """

    #iterate all sets of types of nodes and create node and interconnect all nodes within the set
    def connect_videos(self):
        for unique_label in self.labels:
            for node_id in self.labels[unique_label]['nodes']:
                self.G.add_edge(self.labels[unique_label]['label_id'], node_id, relationship='video-to-label')
                for other_id in self.labels[unique_label]['nodes']:
                    if node_id != other_id:
                        self.G.add_edge(other_id, node_id, relationship='video-to-video')

    #iterate all csv connections and create virtual node connections
    def connect_labels(self, csv_path, include_is_a=True):
        #df_edges = pd.read_csv(csv_path, usecols=['from_action_name', 'to_action_name', 'relation'])
        df_edges = pd.read_csv(csv_path)
        for _, edge_data in tqdm.tqdm(df_edges.iterrows()):
            relation = edge_data['relation']

            #is_a_labels = related_labels(metavd_df, true_label, dataset, 'is-a')
            #similar_labels = related_labels(metavd_df, true_label, dataset, 'similar')
            #equal_labels = related_labels(metavd_df, true_label, dataset, 'equal')

            if relation != 'is-a' or include_is_a:
                action_name_1 = f'{edge_data["from_action_name"]}-{edge_data["from_dataset"]}'
                action_name_2 = f'{edge_data["to_action_name"]}-{edge_data["to_dataset"]}'
                #action_name_1 = f'{edge_data["from_action_name"]}'
                #action_name_2 = f'{edge_data["to_action_name"]}'
                if action_name_1 in self.labels and action_name_2 in self.labels:
                    from_id = self.labels[action_name_1]['label_id']
                    to_id = self.labels[action_name_2]['label_id']
                    self.G.add_edge(from_id, to_id, relationship='label-to-label-'+relation)
                else:
                    if self.verbose:
                        if action_name_1 not in self.labels:
                            print(f'{action_name_1} not in self.labels')
                        if action_name_2 not in self.labels:
                            print(f'{action_name_2} not in self.labels')

    def create_graph(self, data_h5, csv, export_path, export=False):
        self.create_labels(data_h5)
        self.connect_videos()
        self.connect_labels(csv, include_is_a=True)
        if export:
            self.export_graph(export_path)

    #draw the graph
    def draw_graph(self):
        nx.draw(self.G, with_labels=True, font_weight='bold', node_size=300, node_color='skyblue', font_color='black', font_size=8)
        plt.show()

    def export_graph(self, export_path):
        G_export = nx.Graph()
        for node, attrs in self.G.nodes(data=True):
            converted_attrs = {key: str(value) for key, value in attrs.items()}
            G_export.add_node(node, **converted_attrs)

         # Export the new graph to GEXF
        nx.write_graphml(G_export, export_path)

    # def import_graph(self, import_path):
    #     G_loaded = nx.read_graphml(import_path)
    #     for node, attrs in G_loaded.nodes(data=True):
    #         for key, value in attrs.items():


#visualize with plt


#TODO: S

if __name__ == "__main__":
    data_h5 = utils.read_h5("frames_data.h5")
    gl = GraphLoader()
    # gl.create_labels(data_h5)
    # gl.connect_videos()
    # gl.connect_labels('metavd_v1.csv')
    gl.create_graph(data_h5=data_h5, csv='metavd_v1.csv', export_path='metavd_v1.graphml')
    gl.draw_graph()