-
Notifications
You must be signed in to change notification settings - Fork 17
/
textvqa.yml
72 lines (72 loc) · 2.76 KB
/
textvqa.yml
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
task_attributes:
vqa:
datasets:
- textvqa
dataset_size_proportional_sampling: true
dataset_attributes:
textvqa:
data_root_dir: ../data
image_depth_first: false
fast_read: false
fast_dir: /data/like/v1
image_features:
train:
- open_images/detectron_fix_100/fc6/train,open_images/resnet152/train
val:
- open_images/detectron_fix_100/fc6/train,open_images/resnet152/train
test:
- open_images/detectron_fix_100/fc6/test,open_images/resnet152/test
imdb_files:
# train:
# - imdb/textvqa_0.5/imdb_textvqa_train.npy
# val:
# - imdb/textvqa_0.5/imdb_textvqa_val.npy
# test:
# - imdb/textvqa_0.5/imdb_textvqa_test.npy
train:
- imdb/textvqa_0.5/imdb_textvqa_train_trimmed.npy
val:
- imdb/textvqa_0.5/imdb_textvqa_val_trimmed.npy
test:
- imdb/textvqa_0.5/imdb_textvqa_test_trimmed.npy
features_max_len: 137
processors:
text_processor:
type: vocab
params:
max_length: 14
vocab:
type: intersected
embedding_name: glove.6B.300d
vocab_file: vocabs/vocabulary_100k.txt
preprocessor:
type: simple_sentence
params: {}
answer_processor:
type: vqa_answer
params:
vocab_file: vocabs/answers_textvqa_8k.txt
preprocessor:
type: simple_word
params: {}
num_answers: 10
context_processor:
type: fasttext
params:
max_length: 50
model_file: .vector_cache/wiki.en.bin
ocr_token_processor:
type: simple_word
params: {}
bbox_processor:
type: bbox
params:
max_length: 50
return_info: true
# Return OCR information
use_ocr: true
# Return spatial information of OCR tokens if present
use_ocr_info: true
training_parameters:
monitored_metric: textvqa_vqa_accuracy
metric_minimize: false