cgo-pytorch/eval_classifier.py at master · Inlinebool/cgo-pytorch · 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
import argparse
import json
import os

import torch
from torch.utils.data import DataLoader
from tqdm import tqdm

from datasets import ClassifierDataset


def init_result(categories):
    eval_result = {}
    for c in categories:
        eval_result[c] = {
            'true_pos': 0,
            'false_pos': 0,
            'true_neg': 0,
            'false_neg': 0,
        }
    return eval_result


def compute_metrics(result):
    for c in result:
        all_pos = result[c]['true_pos'] + result[c]['false_pos']
        all_relevant = result[c]['true_pos'] + result[c]['false_neg']
        result[c]['recall'] = result[c][
            'true_pos'] / all_relevant if all_relevant else 0
        result[c][
            'precision'] = result[c]['true_pos'] / all_pos if all_pos else 0
        result[c]['f1'] = 2 * result[c]['true_pos'] / (
            2 * result[c]['true_pos'] + result[c]['false_pos'] +
            result[c]['false_neg'])


def hit_fn_topk(k: int):
    return lambda predict: torch.argsort(predict, descending=True)[0:k].view(
        -1).tolist()


def hit_fn_thresh(thresh: float):
    return lambda predict: torch.nonzero(torch.ge(predict, thresh)).view(
        -1).tolist()


def hit_to_names(hit, categories):
    names = [categories[x] for x in hit]
    return names


def evaluate_model(model, test, category_names, hit_fn):
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    model = model.to(device)
    with torch.no_grad():
        data_loader = DataLoader(test, shuffle=True)
        eval_result = init_result(category_names)
        model.eval()
        for _, batch in tqdm(enumerate(data_loader), total=len(data_loader)):
            inputs = [x.to(device) for x in batch['inputs']]
            label = [x.to(device) for x in batch['label']]
            predict = model(inputs)
            predict = torch.sigmoid(predict).squeeze()
            hit = hit_fn(predict)
            # label_hit = hit_fn_thresh(0.9)(label)
            # predicted_categories = hit_to_names(hit, category_names)
            # label_categories = hit_to_names(label_hit, category_names)
            for c in category_names:
                if label[category_names.index(c)] == 1:
                    if category_names.index(c) in hit:
                        eval_result[c]['true_pos'] += 1
                    else:
                        eval_result[c]['false_neg'] += 1
                else:
                    if category_names.index(c) in hit:
                        eval_result[c]['false_pos'] += 1
                    else:
                        eval_result[c]['true_neg'] += 1

        compute_metrics(eval_result)
        return eval_result


if __name__ == "__main__":
    parser = argparse.ArgumentParser(description='Evaluate classifier.')

    parser.add_argument('--model_dir', type=str, default='models')
    parser.add_argument('--model_name', type=str, default='classifier')
    parser.add_argument('--data_dir', type=str, default='data')
    parser.add_argument('--classifier_data_dir',
                        type=str,
                        default='classifier')
    parser.add_argument('--feature_filename',
                        type=str,
                        default='image_features.h5')
    parser.add_argument('--featuremap_filename',
                        type=str,
                        default='feature_map.json')
    parser.add_argument('--label_filename',
                        type=str,
                        default='label_detection.json')
    parser.add_argument('--category_name_filename',
                        type=str,
                        default='category_names.json')
    parser.add_argument('--result_dir', type=str, default='results')
    parser.add_argument('--thresh', type=float, default=0.5)
    parser.add_argument('--top_k', type=int, default=5)

    args = parser.parse_args()

    model_path = os.path.join(args.model_dir, args.model_name)
    feature_path = os.path.join(args.data_dir, args.feature_filename)
    featuremap_path = os.path.join(args.data_dir, args.featuremap_filename)
    label_path = os.path.join(args.data_dir, args.classifier_data_dir,
                              args.label_filename)
    category_name_path = os.path.join(args.data_dir, args.classifier_data_dir,
                                      args.category_name_filename)

    if not os.path.exists(args.result_dir):
        os.mkdir(args.result_dir)

    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

    model = torch.load('{0}.pkl'.format(model_path)).to(device)
    test = ClassifierDataset(feature_path, label_path, featuremap_path, 'test')
    with open(category_name_path, 'r') as fp:
        category_names = json.load(fp)

    results_topk = evaluate_model(model, test, category_names,
                                  hit_fn_topk(args.top_k))
    topk_filename = 'results_{0}_top{1}.json'.format(args.model_name,
                                                     args.top_k)
    with open(os.path.join(args.result_dir, topk_filename), 'w') as fp:
        json.dump(results_topk, fp, indent=4)

    results_thresh = evaluate_model(model, test, category_names,
                                    hit_fn_thresh(args.thresh))
    thresh_filename = 'results_{0}_thresh{1}.json'.format(
        args.model_name, args.thresh)
    with open(os.path.join(args.result_dir, thresh_filename), 'w') as fp:
        json.dump(results_thresh, fp, indent=4)