Global:
device: gpu
epoch_num: 100
log_smooth_window: 20
print_batch_step: 10
output_dir: ./output/rec/repsvtr_test/
save_epoch_step: [150, 10]
# evaluation is run every 2000 iterations
eval_epoch_step: [0, 1]
eval_batch_step: [0, 2000]
cal_metric_during_train: True
pretrained_model: /scratch/penghao/OpenOCR/openocr_repsvtr_ch.pth
checkpoints:
use_tensorboard: false
infer_img: /scratch/penghao/datasets/Date-Real/eval/images
# for data or label process
character_dict_path: &character_dict_path ./tools/utils/ppocr_keys_v1.txt
max_text_length: &max_text_length 25
use_space_char: &use_space_char True
save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_ctc.txt
use_amp: True
project_name: resvtr_ctc_nosgm_ds
Optimizer:
name: Adam
lr: 0.0001
weight_decay: 3.0e-05
filter_bias_and_bn: False
LRScheduler:
name: CosineAnnealingLR
warmup_epoch: 5
Architecture:
model_type: rec
algorithm: SVTRv2_mobile
Transform:
Encoder:
name: RepSVTREncoder
Decoder:
name: CTCDecoder
svtr_encoder:
dims: 256
depth: 2
hidden_dims: 256
kernel_size: [1, 3]
use_guide: True
Loss:
name: CTCLoss
zero_infinity: True
PostProcess:
name: CTCLabelDecode
character_dict_path: *character_dict_path
Metric:
name: RecMetric
main_indicator: acc
ignore_space: False
# is_filter: True
Train:
dataset:
name: SimpleDataSet
data_dir: /scratch/penghao/datasets/Date-Real/train/images
label_file_list:
- /scratch/penghao/datasets/Date-Real/train/labels.txt
transforms:
- DecodeImagePIL: # load image
img_mode: RGB
- PARSeqAugPIL:
- CTCLabelEncode: # Class handling label
character_dict_path: *character_dict_path
use_space_char: *use_space_char
max_text_length: *max_text_length
- RecTVResize:
image_shape: [48, 320]
padding: True
- KeepKeys:
keep_keys: ['image', 'label', 'length']
loader:
shuffle: True
batch_size_per_card: 256
drop_last: True
num_workers: 8
Eval:
dataset:
name: SimpleDataSet
data_dir: /scratch/penghao/datasets/Date-Real/train/images
label_file_list:
- /scratch/penghao/datasets/Date-Real/train/labels.txt
transforms:
- DecodeImagePIL: # load image
img_mode: RGB
- CTCLabelEncode: # Class handling label
character_dict_path: *character_dict_path
use_space_char: *use_space_char
max_text_length: *max_text_length
- RecTVResize:
image_shape: [48, 320]
padding: True
# - SVTRResize:
# image_shape: [3, 48, 320]
# padding: True
- KeepKeys:
keep_keys: ['image', 'label', 'length']
loader:
shuffle: False
drop_last: False
batch_size_per_card: 1
num_workers: 8
