2022-05-22 13:16:52 +08:00
|
|
|
Global:
|
|
|
|
|
use_gpu: true
|
|
|
|
|
epoch_num: 5
|
|
|
|
|
log_smooth_window: 20
|
|
|
|
|
print_batch_step: 20
|
|
|
|
|
save_model_dir: ./output/rec/rec_r31_robustscanner/
|
|
|
|
|
save_epoch_step: 1
|
|
|
|
|
# evaluation is run every 2000 iterations
|
|
|
|
|
eval_batch_step: [0, 2000]
|
|
|
|
|
cal_metric_during_train: True
|
|
|
|
|
pretrained_model:
|
|
|
|
|
checkpoints:
|
|
|
|
|
save_inference_dir:
|
|
|
|
|
use_visualdl: False
|
2022-09-29 14:52:10 +08:00
|
|
|
infer_img: doc/imgs_words_en/word_10.png
|
2022-05-22 13:16:52 +08:00
|
|
|
# for data or label process
|
|
|
|
|
character_dict_path: ppocr/utils/dict90.txt
|
2022-06-12 12:25:34 +08:00
|
|
|
max_text_length: &max_text_length 40
|
2022-05-22 13:16:52 +08:00
|
|
|
infer_mode: False
|
|
|
|
|
use_space_char: False
|
|
|
|
|
rm_symbol: True
|
|
|
|
|
save_res_path: ./output/rec/predicts_robustscanner.txt
|
|
|
|
|
|
|
|
|
|
Optimizer:
|
|
|
|
|
name: Adam
|
|
|
|
|
beta1: 0.9
|
|
|
|
|
beta2: 0.999
|
|
|
|
|
lr:
|
|
|
|
|
name: Piecewise
|
|
|
|
|
decay_epochs: [3, 4]
|
|
|
|
|
values: [0.001, 0.0001, 0.00001]
|
|
|
|
|
regularizer:
|
|
|
|
|
name: 'L2'
|
|
|
|
|
factor: 0
|
|
|
|
|
|
|
|
|
|
Architecture:
|
|
|
|
|
model_type: rec
|
|
|
|
|
algorithm: RobustScanner
|
|
|
|
|
Transform:
|
|
|
|
|
Backbone:
|
2022-06-12 12:25:34 +08:00
|
|
|
name: ResNet31
|
2022-07-14 23:11:20 +08:00
|
|
|
init_type: KaimingNormal
|
2022-05-22 13:16:52 +08:00
|
|
|
Head:
|
|
|
|
|
name: RobustScannerHead
|
|
|
|
|
enc_outchannles: 128
|
|
|
|
|
hybrid_dec_rnn_layers: 2
|
|
|
|
|
hybrid_dec_dropout: 0
|
|
|
|
|
position_dec_rnn_layers: 2
|
|
|
|
|
start_idx: 91
|
|
|
|
|
mask: True
|
|
|
|
|
padding_idx: 92
|
|
|
|
|
encode_value: False
|
2022-06-12 12:25:34 +08:00
|
|
|
max_text_length: *max_text_length
|
2022-05-22 13:16:52 +08:00
|
|
|
|
|
|
|
|
Loss:
|
|
|
|
|
name: SARLoss
|
|
|
|
|
|
|
|
|
|
PostProcess:
|
|
|
|
|
name: SARLabelDecode
|
|
|
|
|
|
|
|
|
|
Metric:
|
|
|
|
|
name: RecMetric
|
|
|
|
|
is_filter: True
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Train:
|
|
|
|
|
dataset:
|
2022-07-14 23:11:20 +08:00
|
|
|
name: LMDBDataSet
|
2022-08-01 22:20:34 +08:00
|
|
|
data_dir: ./train_data/data_lmdb_release/training/
|
2022-05-22 13:16:52 +08:00
|
|
|
transforms:
|
|
|
|
|
- DecodeImage: # load image
|
|
|
|
|
img_mode: BGR
|
|
|
|
|
channel_first: False
|
|
|
|
|
- SARLabelEncode: # Class handling label
|
|
|
|
|
- RobustScannerRecResizeImg:
|
|
|
|
|
image_shape: [3, 48, 48, 160] # h:48 w:[48,160]
|
|
|
|
|
width_downsample_ratio: 0.25
|
2022-06-12 12:25:34 +08:00
|
|
|
max_text_length: *max_text_length
|
2022-05-22 13:16:52 +08:00
|
|
|
- KeepKeys:
|
|
|
|
|
keep_keys: ['image', 'label', 'valid_ratio', 'word_positons'] # dataloader will return list in this order
|
|
|
|
|
loader:
|
|
|
|
|
shuffle: True
|
2022-06-12 12:25:34 +08:00
|
|
|
batch_size_per_card: 64
|
2022-05-22 13:16:52 +08:00
|
|
|
drop_last: True
|
2022-06-12 12:25:34 +08:00
|
|
|
num_workers: 8
|
2022-05-22 13:16:52 +08:00
|
|
|
use_shared_memory: False
|
|
|
|
|
|
|
|
|
|
Eval:
|
|
|
|
|
dataset:
|
|
|
|
|
name: LMDBDataSet
|
2022-08-01 22:20:34 +08:00
|
|
|
data_dir: ./train_data/data_lmdb_release/evaluation/
|
2022-05-22 13:16:52 +08:00
|
|
|
transforms:
|
|
|
|
|
- DecodeImage: # load image
|
|
|
|
|
img_mode: BGR
|
|
|
|
|
channel_first: False
|
|
|
|
|
- SARLabelEncode: # Class handling label
|
|
|
|
|
- RobustScannerRecResizeImg:
|
|
|
|
|
image_shape: [3, 48, 48, 160]
|
2022-07-11 23:28:51 +08:00
|
|
|
max_text_length: *max_text_length
|
2022-05-22 13:16:52 +08:00
|
|
|
width_downsample_ratio: 0.25
|
|
|
|
|
- KeepKeys:
|
|
|
|
|
keep_keys: ['image', 'label', 'valid_ratio', 'word_positons'] # dataloader will return list in this order
|
|
|
|
|
loader:
|
|
|
|
|
shuffle: False
|
|
|
|
|
drop_last: False
|
2022-06-12 12:25:34 +08:00
|
|
|
batch_size_per_card: 64
|
|
|
|
|
num_workers: 4
|
2022-05-22 13:16:52 +08:00
|
|
|
use_shared_memory: False
|
|
|
|
|
|
