Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
聚火科技码农
PaddleOCR
提交
b722eb56
P
PaddleOCR
项目概览
聚火科技码农
/
PaddleOCR
与 Fork 源项目一致
Fork自
PaddlePaddle / PaddleOCR
通知
1
Star
0
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
0
列表
看板
标记
里程碑
合并请求
0
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
PaddleOCR
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
0
Issue
0
列表
看板
标记
里程碑
合并请求
0
合并请求
0
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
提交
b722eb56
编写于
6月 03, 2020
作者:
T
tink2123
浏览文件
操作
浏览文件
下载
差异文件
fix infer_rec for attention
上级
b4c5dac2
ade18e13
变更
17
隐藏空白更改
内联
并排
Showing
17 changed file
with
108 addition
and
49 deletion
+108
-49
configs/rec/rec_chinese_lite_train.yml
configs/rec/rec_chinese_lite_train.yml
+3
-2
configs/rec/rec_icdar15_train.yml
configs/rec/rec_icdar15_train.yml
+2
-1
configs/rec/rec_mv3_none_bilstm_ctc.yml
configs/rec/rec_mv3_none_bilstm_ctc.yml
+3
-2
configs/rec/rec_mv3_none_none_ctc.yml
configs/rec/rec_mv3_none_none_ctc.yml
+1
-0
configs/rec/rec_mv3_tps_bilstm_attn.yml
configs/rec/rec_mv3_tps_bilstm_attn.yml
+2
-1
configs/rec/rec_mv3_tps_bilstm_ctc.yml
configs/rec/rec_mv3_tps_bilstm_ctc.yml
+1
-0
configs/rec/rec_r34_vd_none_bilstm_ctc.yml
configs/rec/rec_r34_vd_none_bilstm_ctc.yml
+1
-0
configs/rec/rec_r34_vd_none_none_ctc.yml
configs/rec/rec_r34_vd_none_none_ctc.yml
+1
-0
configs/rec/rec_r34_vd_tps_bilstm_attn.yml
configs/rec/rec_r34_vd_tps_bilstm_attn.yml
+1
-0
configs/rec/rec_r34_vd_tps_bilstm_ctc.yml
configs/rec/rec_r34_vd_tps_bilstm_ctc.yml
+1
-0
ppocr/data/det/db_process.py
ppocr/data/det/db_process.py
+3
-0
ppocr/data/rec/dataset_traversal.py
ppocr/data/rec/dataset_traversal.py
+10
-6
ppocr/data/rec/img_tools.py
ppocr/data/rec/img_tools.py
+1
-1
ppocr/modeling/architectures/rec_model.py
ppocr/modeling/architectures/rec_model.py
+6
-2
ppocr/modeling/heads/rec_attention_head.py
ppocr/modeling/heads/rec_attention_head.py
+9
-4
tools/infer/predict_rec.py
tools/infer/predict_rec.py
+44
-21
tools/infer_rec.py
tools/infer_rec.py
+19
-9
未找到文件。
configs/rec/rec_chinese_lite_train.yml
浏览文件 @
b722eb56
Global
:
algorithm
:
CRNN
use_gpu
:
tru
e
use_gpu
:
fals
e
epoch_num
:
3000
log_smooth_window
:
20
print_batch_step
:
10
...
...
@@ -8,6 +8,7 @@ Global:
save_epoch_step
:
3
eval_batch_step
:
2000
train_batch_size_per_card
:
256
drop_last
:
true
test_batch_size_per_card
:
256
image_shape
:
[
3
,
32
,
320
]
max_text_length
:
25
...
...
@@ -15,7 +16,7 @@ Global:
character_dict_path
:
./ppocr/utils/ppocr_keys_v1.txt
loss_type
:
ctc
reader_yml
:
./configs/rec/rec_chinese_reader.yml
pretrain_weights
:
pretrain_weights
:
output/rec_CRNN/rec_mv3_crnn/best_accuracy
checkpoints
:
save_inference_dir
:
infer_img
:
...
...
configs/rec/rec_icdar15_train.yml
浏览文件 @
b722eb56
...
...
@@ -8,13 +8,14 @@ Global:
save_epoch_step
:
300
eval_batch_step
:
500
train_batch_size_per_card
:
256
drop_last
:
true
test_batch_size_per_card
:
256
image_shape
:
[
3
,
32
,
100
]
max_text_length
:
25
character_type
:
en
loss_type
:
ctc
reader_yml
:
./configs/rec/rec_icdar15_reader.yml
pretrain_weights
:
./pretrain_models/rec_mv3_none_bilstm_ctc/best_accuracy
pretrain_weights
:
checkpoints
:
save_inference_dir
:
infer_img
:
...
...
configs/rec/rec_mv3_none_bilstm_ctc.yml
浏览文件 @
b722eb56
Global
:
algorithm
:
CRNN
use_gpu
:
tru
e
use_gpu
:
fals
e
epoch_num
:
72
log_smooth_window
:
20
print_batch_step
:
10
...
...
@@ -8,13 +8,14 @@ Global:
save_epoch_step
:
3
eval_batch_step
:
2000
train_batch_size_per_card
:
256
drop_last
:
true
test_batch_size_per_card
:
256
image_shape
:
[
3
,
32
,
100
]
max_text_length
:
25
character_type
:
en
loss_type
:
ctc
reader_yml
:
./configs/rec/rec_benchmark_reader.yml
pretrain_weights
:
./output/rec_CRNN/rec_mv3_none_bilstm_ctc/best_accuracy
pretrain_weights
:
checkpoints
:
save_inference_dir
:
infer_img
:
...
...
configs/rec/rec_mv3_none_none_ctc.yml
浏览文件 @
b722eb56
...
...
@@ -8,6 +8,7 @@ Global:
save_epoch_step
:
3
eval_batch_step
:
2000
train_batch_size_per_card
:
256
drop_last
:
true
test_batch_size_per_card
:
256
image_shape
:
[
3
,
32
,
100
]
max_text_length
:
25
...
...
configs/rec/rec_mv3_tps_bilstm_attn.yml
浏览文件 @
b722eb56
Global
:
algorithm
:
RARE
use_gpu
:
tru
e
use_gpu
:
fals
e
epoch_num
:
72
log_smooth_window
:
20
print_batch_step
:
10
...
...
@@ -8,6 +8,7 @@ Global:
save_epoch_step
:
3
eval_batch_step
:
2000
train_batch_size_per_card
:
256
drop_last
:
true
test_batch_size_per_card
:
256
image_shape
:
[
3
,
32
,
100
]
max_text_length
:
25
...
...
configs/rec/rec_mv3_tps_bilstm_ctc.yml
浏览文件 @
b722eb56
...
...
@@ -8,6 +8,7 @@ Global:
save_epoch_step
:
3
eval_batch_step
:
2000
train_batch_size_per_card
:
256
drop_last
:
true
test_batch_size_per_card
:
256
image_shape
:
[
3
,
32
,
100
]
max_text_length
:
25
...
...
configs/rec/rec_r34_vd_none_bilstm_ctc.yml
浏览文件 @
b722eb56
...
...
@@ -8,6 +8,7 @@ Global:
save_epoch_step
:
3
eval_batch_step
:
2000
train_batch_size_per_card
:
256
drop_last
:
true
test_batch_size_per_card
:
256
image_shape
:
[
3
,
32
,
100
]
max_text_length
:
25
...
...
configs/rec/rec_r34_vd_none_none_ctc.yml
浏览文件 @
b722eb56
...
...
@@ -8,6 +8,7 @@ Global:
save_epoch_step
:
3
eval_batch_step
:
2000
train_batch_size_per_card
:
256
drop_last
:
true
test_batch_size_per_card
:
256
image_shape
:
[
3
,
32
,
100
]
max_text_length
:
25
...
...
configs/rec/rec_r34_vd_tps_bilstm_attn.yml
浏览文件 @
b722eb56
...
...
@@ -8,6 +8,7 @@ Global:
save_epoch_step
:
3
eval_batch_step
:
2000
train_batch_size_per_card
:
256
drop_last
:
true
test_batch_size_per_card
:
256
image_shape
:
[
3
,
32
,
100
]
max_text_length
:
25
...
...
configs/rec/rec_r34_vd_tps_bilstm_ctc.yml
浏览文件 @
b722eb56
...
...
@@ -8,6 +8,7 @@ Global:
save_epoch_step
:
3
eval_batch_step
:
2000
train_batch_size_per_card
:
256
drop_last
:
true
test_batch_size_per_card
:
256
image_shape
:
[
3
,
32
,
100
]
max_text_length
:
25
...
...
ppocr/data/det/db_process.py
浏览文件 @
b722eb56
...
...
@@ -17,6 +17,8 @@ import cv2
import
numpy
as
np
import
json
import
sys
from
ppocr.utils.utility
import
initial_logger
logger
=
initial_logger
()
from
.data_augment
import
AugmentData
from
.random_crop_data
import
RandomCropData
...
...
@@ -100,6 +102,7 @@ class DBProcessTrain(object):
img_path
,
gt_label
=
self
.
convert_label_infor
(
label_infor
)
imgvalue
=
cv2
.
imread
(
img_path
)
if
imgvalue
is
None
:
logger
.
info
(
"{} does not exist!"
.
format
(
img_path
))
return
None
data
=
self
.
make_data_dict
(
imgvalue
,
gt_label
)
data
=
AugmentData
(
data
)
...
...
ppocr/data/rec/dataset_traversal.py
浏览文件 @
b722eb56
...
...
@@ -43,6 +43,7 @@ class LMDBReader(object):
self
.
mode
=
params
[
'mode'
]
if
params
[
'mode'
]
==
'train'
:
self
.
batch_size
=
params
[
'train_batch_size_per_card'
]
self
.
drop_last
=
params
[
'drop_last'
]
else
:
self
.
batch_size
=
params
[
'test_batch_size_per_card'
]
self
.
infer_img
=
params
[
'infer_img'
]
...
...
@@ -99,7 +100,7 @@ class LMDBReader(object):
process_id
=
0
def
sample_iter_reader
():
if
self
.
infer_img
is
not
None
:
if
self
.
mode
!=
'train'
and
self
.
infer_img
is
not
None
:
image_file_list
=
get_image_file_list
(
self
.
infer_img
)
for
single_img
in
image_file_list
:
img
=
cv2
.
imread
(
single_img
)
...
...
@@ -146,10 +147,11 @@ class LMDBReader(object):
if
len
(
batch_outs
)
==
self
.
batch_size
:
yield
batch_outs
batch_outs
=
[]
if
len
(
batch_outs
)
!=
0
:
yield
batch_outs
if
not
self
.
drop_last
:
if
len
(
batch_outs
)
!=
0
:
yield
batch_outs
if
self
.
infer_img
is
None
:
if
self
.
mode
!=
'train'
and
self
.
infer_img
is
None
:
return
batch_iter_reader
return
sample_iter_reader
...
...
@@ -171,6 +173,7 @@ class SimpleReader(object):
self
.
infer_img
=
params
[
'infer_img'
]
if
params
[
'mode'
]
==
'train'
:
self
.
batch_size
=
params
[
'train_batch_size_per_card'
]
self
.
drop_last
=
params
[
'drop_last'
]
else
:
self
.
batch_size
=
params
[
'test_batch_size_per_card'
]
...
...
@@ -226,8 +229,9 @@ class SimpleReader(object):
if
len
(
batch_outs
)
==
self
.
batch_size
:
yield
batch_outs
batch_outs
=
[]
if
len
(
batch_outs
)
!=
0
:
yield
batch_outs
if
not
self
.
drop_last
:
if
len
(
batch_outs
)
!=
0
:
yield
batch_outs
if
self
.
infer_img
is
None
:
return
batch_iter_reader
...
...
ppocr/data/rec/img_tools.py
浏览文件 @
b722eb56
...
...
@@ -51,7 +51,7 @@ def resize_norm_img(img, image_shape):
def
resize_norm_img_chinese
(
img
,
image_shape
):
imgC
,
imgH
,
imgW
=
image_shape
# todo: change to 0 and modified image shape
max_wh_ratio
=
1
0
max_wh_ratio
=
0
h
,
w
=
img
.
shape
[
0
],
img
.
shape
[
1
]
ratio
=
w
*
1.0
/
h
max_wh_ratio
=
max
(
max_wh_ratio
,
ratio
)
...
...
ppocr/modeling/architectures/rec_model.py
浏览文件 @
b722eb56
...
...
@@ -110,7 +110,11 @@ class RecModel(object):
return
loader
,
outputs
elif
mode
==
"export"
:
predict
=
predicts
[
'predict'
]
predict
=
fluid
.
layers
.
softmax
(
predict
)
if
self
.
loss_type
==
"ctc"
:
predict
=
fluid
.
layers
.
softmax
(
predict
)
return
[
image
,
{
'decoded_out'
:
decoded_out
,
'predicts'
:
predict
}]
else
:
return
loader
,
{
'decoded_out'
:
decoded_out
}
predict
=
predicts
[
'predict'
]
if
self
.
loss_type
==
"ctc"
:
predict
=
fluid
.
layers
.
softmax
(
predict
)
return
loader
,
{
'decoded_out'
:
decoded_out
,
'predicts'
:
predict
}
ppocr/modeling/heads/rec_attention_head.py
浏览文件 @
b722eb56
...
...
@@ -123,6 +123,8 @@ class AttentionPredict(object):
full_ids
=
fluid
.
layers
.
fill_constant_batch_size_like
(
input
=
init_state
,
shape
=
[
-
1
,
1
],
dtype
=
'int64'
,
value
=
1
)
full_scores
=
fluid
.
layers
.
fill_constant_batch_size_like
(
input
=
init_state
,
shape
=
[
-
1
,
1
],
dtype
=
'float32'
,
value
=
1
)
cond
=
layers
.
less_than
(
x
=
counter
,
y
=
array_len
)
while_op
=
layers
.
While
(
cond
=
cond
)
...
...
@@ -171,6 +173,9 @@ class AttentionPredict(object):
new_ids
=
fluid
.
layers
.
concat
([
full_ids
,
topk_indices
],
axis
=
1
)
fluid
.
layers
.
assign
(
new_ids
,
full_ids
)
new_scores
=
fluid
.
layers
.
concat
([
full_scores
,
topk_scores
],
axis
=
1
)
fluid
.
layers
.
assign
(
new_scores
,
full_scores
)
layers
.
increment
(
x
=
counter
,
value
=
1
,
in_place
=
True
)
# update the memories
...
...
@@ -184,7 +189,7 @@ class AttentionPredict(object):
length_cond
=
layers
.
less_than
(
x
=
counter
,
y
=
array_len
)
finish_cond
=
layers
.
logical_not
(
layers
.
is_empty
(
x
=
topk_indices
))
layers
.
logical_and
(
x
=
length_cond
,
y
=
finish_cond
,
out
=
cond
)
return
full_ids
return
full_ids
,
full_scores
def
__call__
(
self
,
inputs
,
labels
=
None
,
mode
=
None
):
encoder_features
=
self
.
encoder
(
inputs
)
...
...
@@ -223,10 +228,10 @@ class AttentionPredict(object):
decoder_size
,
char_num
)
_
,
decoded_out
=
layers
.
topk
(
input
=
predict
,
k
=
1
)
decoded_out
=
layers
.
lod_reset
(
decoded_out
,
y
=
label_out
)
predicts
=
{
'predict'
:
predict
,
'decoded_out'
:
decoded_out
}
predicts
=
{
'predict'
:
predict
,
'decoded_out'
:
decoded_out
}
else
:
ids
=
self
.
gru_attention_infer
(
ids
,
predict
=
self
.
gru_attention_infer
(
decoder_boot
,
self
.
max_length
,
char_num
,
word_vector_dim
,
encoded_vector
,
encoded_proj
,
decoder_size
)
predicts
=
{
'
decoded_out'
:
ids
}
predicts
=
{
'
predict'
:
predict
,
'decoded_out'
:
ids
}
return
predicts
tools/infer/predict_rec.py
浏览文件 @
b722eb56
...
...
@@ -80,26 +80,43 @@ class TextRecognizer(object):
starttime
=
time
.
time
()
self
.
input_tensor
.
copy_from_cpu
(
norm_img_batch
)
self
.
predictor
.
zero_copy_run
()
rec_idx_batch
=
self
.
output_tensors
[
0
].
copy_to_cpu
()
rec_idx_lod
=
self
.
output_tensors
[
0
].
lod
()[
0
]
predict_batch
=
self
.
output_tensors
[
1
].
copy_to_cpu
()
predict_lod
=
self
.
output_tensors
[
1
].
lod
()[
0
]
elapse
=
time
.
time
()
-
starttime
predict_time
+=
elapse
starttime
=
time
.
time
()
for
rno
in
range
(
len
(
rec_idx_lod
)
-
1
):
beg
=
rec_idx_lod
[
rno
]
end
=
rec_idx_lod
[
rno
+
1
]
rec_idx_tmp
=
rec_idx_batch
[
beg
:
end
,
0
]
preds_text
=
self
.
char_ops
.
decode
(
rec_idx_tmp
)
beg
=
predict_lod
[
rno
]
end
=
predict_lod
[
rno
+
1
]
probs
=
predict_batch
[
beg
:
end
,
:]
ind
=
np
.
argmax
(
probs
,
axis
=
1
)
blank
=
probs
.
shape
[
1
]
valid_ind
=
np
.
where
(
ind
!=
(
blank
-
1
))[
0
]
score
=
np
.
mean
(
probs
[
valid_ind
,
ind
[
valid_ind
]])
rec_res
.
append
([
preds_text
,
score
])
if
args
.
rec_algorithm
!=
"RARE"
:
rec_idx_batch
=
self
.
output_tensors
[
0
].
copy_to_cpu
()
rec_idx_lod
=
self
.
output_tensors
[
0
].
lod
()[
0
]
predict_batch
=
self
.
output_tensors
[
1
].
copy_to_cpu
()
predict_lod
=
self
.
output_tensors
[
1
].
lod
()[
0
]
elapse
=
time
.
time
()
-
starttime
predict_time
+=
elapse
for
rno
in
range
(
len
(
rec_idx_lod
)
-
1
):
beg
=
rec_idx_lod
[
rno
]
end
=
rec_idx_lod
[
rno
+
1
]
rec_idx_tmp
=
rec_idx_batch
[
beg
:
end
,
0
]
preds_text
=
self
.
char_ops
.
decode
(
rec_idx_tmp
)
beg
=
predict_lod
[
rno
]
end
=
predict_lod
[
rno
+
1
]
probs
=
predict_batch
[
beg
:
end
,
:]
ind
=
np
.
argmax
(
probs
,
axis
=
1
)
blank
=
probs
.
shape
[
1
]
valid_ind
=
np
.
where
(
ind
!=
(
blank
-
1
))[
0
]
score
=
np
.
mean
(
probs
[
valid_ind
,
ind
[
valid_ind
]])
rec_res
.
append
([
preds_text
,
score
])
else
:
rec_idx_batch
=
self
.
output_tensors
[
0
].
copy_to_cpu
()
predict_batch
=
self
.
output_tensors
[
1
].
copy_to_cpu
()
for
rno
in
range
(
len
(
rec_idx_batch
)):
end_pos
=
np
.
where
(
rec_idx_batch
[
rno
,
:]
==
1
)[
0
]
if
len
(
end_pos
)
<=
1
:
preds
=
rec_idx_batch
[
rno
,
1
:]
score
=
np
.
mean
(
predict_batch
[
rno
,
1
:])
else
:
preds
=
rec_idx_batch
[
rno
,
1
:
end_pos
[
1
]]
score
=
np
.
mean
(
predict_batch
[
rno
,
1
:
end_pos
[
1
]])
#todo: why index has 2 offset
preds
=
preds
-
2
preds_text
=
self
.
char_ops
.
decode
(
preds
)
rec_res
.
append
([
preds_text
,
score
])
return
rec_res
,
predict_time
...
...
@@ -116,7 +133,13 @@ if __name__ == "__main__":
continue
valid_image_file_list
.
append
(
image_file
)
img_list
.
append
(
img
)
rec_res
,
predict_time
=
text_recognizer
(
img_list
)
try
:
rec_res
,
predict_time
=
text_recognizer
(
img_list
)
except
:
logger
.
info
(
"ERROR!!
\n
Input image shape is not equal with config. TPS does not support variable shape.
\n
"
"Please set --rec_image_shape=input_shape and --rec_char_type='ch' "
)
exit
()
for
ino
in
range
(
len
(
img_list
)):
print
(
"Predicts of %s:%s"
%
(
valid_image_file_list
[
ino
],
rec_res
[
ino
]))
print
(
"Total predict time for %d images:%.3f"
%
...
...
tools/infer_rec.py
浏览文件 @
b722eb56
...
...
@@ -55,6 +55,7 @@ def main():
program
.
merge_config
(
FLAGS
.
opt
)
logger
.
info
(
config
)
char_ops
=
CharacterOps
(
config
[
'Global'
])
loss_type
=
config
[
'Global'
][
'loss_type'
]
config
[
'Global'
][
'char_ops'
]
=
char_ops
# check if set use_gpu=True in paddlepaddle cpu version
...
...
@@ -85,29 +86,38 @@ def main():
if
len
(
infer_list
)
==
0
:
logger
.
info
(
"Can not find img in infer_img dir."
)
for
i
in
range
(
max_img_num
):
print
(
"infer_img:
"
,
infer_list
[
i
])
print
(
"infer_img:
%s"
%
infer_list
[
i
])
img
=
next
(
blobs
)
predict
=
exe
.
run
(
program
=
eval_prog
,
feed
=
{
"image"
:
img
},
fetch_list
=
fetch_varname_list
,
return_numpy
=
False
)
preds
=
np
.
array
(
predict
[
0
])
if
preds
.
shape
[
1
]
==
1
:
if
loss_type
==
"ctc"
:
preds
=
np
.
array
(
predict
[
0
])
preds
=
preds
.
reshape
(
-
1
)
preds_lod
=
predict
[
0
].
lod
()[
0
]
preds_text
=
char_ops
.
decode
(
preds
)
else
:
probs
=
np
.
array
(
predict
[
1
])
ind
=
np
.
argmax
(
probs
,
axis
=
1
)
blank
=
probs
.
shape
[
1
]
valid_ind
=
np
.
where
(
ind
!=
(
blank
-
1
))[
0
]
score
=
np
.
mean
(
probs
[
valid_ind
,
ind
[
valid_ind
]])
elif
loss_type
==
"attention"
:
preds
=
np
.
array
(
predict
[
0
])
probs
=
np
.
array
(
predict
[
1
])
end_pos
=
np
.
where
(
preds
[
0
,
:]
==
1
)[
0
]
if
len
(
end_pos
)
<=
1
:
preds_text
=
preds
[
0
,
1
:]
preds
=
preds
[
0
,
1
:]
score
=
np
.
mean
(
probs
[
0
,
1
:])
else
:
preds_text
=
preds
[
0
,
1
:
end_pos
[
1
]]
preds_text
=
preds_text
.
reshape
(
-
1
)
preds_text
=
char_ops
.
decode
(
preds_text
)
preds
=
preds
[
0
,
1
:
end_pos
[
1
]]
score
=
np
.
mean
(
probs
[
0
,
1
:
end_pos
[
1
]])
preds
=
preds
.
reshape
(
-
1
)
preds_text
=
char_ops
.
decode
(
preds
)
print
(
"
\t
index:"
,
preds
)
print
(
"
\t
word :"
,
preds_text
)
print
(
"
\t
score :"
,
score
)
# save for inference model
target_var
=
[]
...
...
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录