-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy patheval_main.py
128 lines (94 loc) · 3.91 KB
/
eval_main.py
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
from lib.data_utils import *
from hallucination_eval.utils.chair import main as chair_main
from hallucination_eval.eval_coco_cap import main as eval_coco_cap_main
from hallucination_eval.ana_hallu_single import main as ana_hallu_single_main
import json
def eval_coco(args):
device = args['device']
save_paths = get_save_paths(args)
output_path = save_paths['output_path']
res = {}
chair_res = chair_main(output_path, args['data_path'] + '/annotations/')
overall_chair_scores = chair_res['overall_metrics']
res['CHAIRs'] = overall_chair_scores['CHAIRs']
res['CHAIRi'] = overall_chair_scores['CHAIRi']
# save chair_res to json
save_res_to_json(chair_res, save_paths['res_intermediate_path'])
coverage_res = ana_hallu_single_main(save_paths['res_intermediate_path'], args)
for k, v in coverage_res.items():
res[k] = v
# based on the chair_res, we can get basic stastic (e.g. avg sentence length) the coverage, and hallucination ratios
coco_eval_res = eval_coco_cap_main(output_path, device=device, data_path=args['data_path'])
for k, v in coco_eval_res.items():
res[k] = v
# save res to json
final_res = {}
final_res['res'] = res
final_res['config'] = args
return final_res
import pandas as pd
from nltk.corpus import wordnet
def get_synonyms(word):
synonyms = set()
for syn in wordnet.synsets(word):
for lemma in syn.lemmas():
synonyms.add(lemma.name())
return list(synonyms)
from data.nocaps_util import CHAIR as NocapsCHAIR
def eval_nocaps(args):
'''
follow https://arxiv.org/pdf/2210.07688.pdf
only add two
types of object categories to our final object list: 1)
super-categories that have sub-categories, and 2)
object categories that have neither super-category
nor sub-categories. Eventually, we construct a list
of 139 coarse-grained object categories from the
600 classes.
https://github.com/nocaps-org/image-feature-extractors/blob/master/data/oi_categories.json
https://storage.googleapis.com/openimages/web/download_v7.html#df-classes-hierarchy
'''
print('evaluating nocaps...')
save_paths = get_save_paths(args)
output_path = save_paths['output_path']
image_ids = []
for item in json.load(open(output_path, 'r')):
image_ids.append(item['image_id'])
chair_main = NocapsCHAIR(image_ids, args)
chair_res = chair_main.compute_chair(output_path)
save_res_to_json(chair_res, save_paths['res_intermediate_path'])
res = {}
overall_chair_scores = chair_res['overall_metrics']
res['CHAIRs'] = overall_chair_scores['CHAIRs']
res['CHAIRi'] = overall_chair_scores['CHAIRi']
print('CHAIRs', res['CHAIRs'])
print('CHAIRi', res['CHAIRi'])
# print('res', res)
basic_res = ana_hallu_single_main(save_paths['res_intermediate_path'], args)
for k, v in basic_res.items():
res[k] = v
# res['sample_num'] = len(image_ids)
# save res to json
final_res = {}
final_res['res'] = res
final_res['config'] = args
return final_res
def main(args, logging=None):
dataset_name = args['dataset_name']
save_paths = get_save_paths(args)
if dataset_name in ['mscoco_captions', 'mscoco']:
# if dataset_name == 'mscoco_captions' or dataset_name == 'mscoco':
final_res = eval_coco(args)
elif dataset_name == 'nocaps':
final_res = eval_nocaps(args)
else:
print('dataset_name {} not supported yet'.format(dataset_name))
raise NotImplementedError
save_res_to_json(final_res, save_paths['res_path'])
print('final res saved to {}'.format(save_paths['res_path']))
if __name__ == '__main__':
args = {
'dataset_name': 'nocaps',
'data_path': '/data/ailin/nocaps/',
}
eval_nocaps(args)