From 8e679d384618c365a45705f04ca2b438b3591174 Mon Sep 17 00:00:00 2001 From: LDOUBLEV Date: Sun, 24 Apr 2022 17:38:43 +0800 Subject: [PATCH] add config --- .../det/ch_PP-OCRv3/ch_PP-OCRv3_det_cml.yml | 239 ++++++++++++++++++ .../ch_PP-OCRv3/ch_PP-OCRv3_det_student.yml | 178 +++++++++++++ 2 files changed, 417 insertions(+) create mode 100644 configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_cml.yml create mode 100644 configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_student.yml diff --git a/configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_cml.yml b/configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_cml.yml new file mode 100644 index 00000000..d8001197 --- /dev/null +++ b/configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_cml.yml @@ -0,0 +1,239 @@ +Global: + debug: false + use_gpu: true + epoch_num: 500 + log_smooth_window: 20 + print_batch_step: 10 + save_model_dir: ./output/ch_PP-OCR_V3/ + save_epoch_step: 100 + eval_batch_step: + - 0 + - 400 + load_static_weights: false + cal_metric_during_train: false + pretrained_model: null + checkpoints: null + save_inference_dir: null + use_visualdl: false + infer_img: doc/imgs_en/img_10.jpg + save_res_path: ./checkpoints/det_db/predicts_db.txt + distributed: true + student_pretrained_model: ./pretrain_model/MobileNetV3_large_x0_5_pretrained + teacher_pretrained_model: ./inference/ch_ppocr_resnet50/student +Architecture: + model_type: det + algorithm: DB +NetWorks: + Student: + model_type: det + algorithm: DB + Transform: null + Backbone: + name: MobileNetV3 + scale: 0.5 + model_name: large + disable_se: true + Neck: + name: FEPAN + out_channels: 96 + Head: + name: DBHead + k: 50 + kernel_size: [7, 2, 2] + Student2: + model_type: det + algorithm: DB + Transform: null + Backbone: + name: MobileNetV3 + scale: 0.5 + model_name: large + disable_se: true + Neck: + name: FEPAN + out_channels: 96 + Head: + name: DBHead + k: 50 + kernel_size: [7, 2, 2] + Teacher: + model_type: det + algorithm: DB + Backbone: + name: ResNet + in_channels: 3 + layers: 50 + Neck: + name: DBFPN + out_channels: 256 + big_kernel: true + Head: + name: DBHead + big_kernel: true + k: 50 + +Loss: + name: CombinedLoss + loss_config_list: + - DistillationDilaDBLoss: + weight: 1.0 + model_name_pairs: + - ["Student", "Teacher"] + - ["Student2", "Teacher"] + key: maps + balance_loss: true + main_loss_type: DiceLoss + alpha: 5 + beta: 10 + ohem_ratio: 3 + - DistillationDMLLoss: + model_name_pairs: + - ["Student", "Student2"] + maps_name: "thrink_maps" + weight: 1.0 + # act: None + model_name_pairs: ["Student", "Student2"] + key: maps + - DistillationDBLoss: + weight: 1.0 + model_name_list: ["Student", "Student2"] + # key: maps + # name: DBLoss + balance_loss: true + main_loss_type: DiceLoss + alpha: 5 + beta: 10 + ohem_ratio: 3 + +Optimizer: + name: Adam + beta1: 0.9 + beta2: 0.999 + lr: + name: Cosine + learning_rate: 0.001 + warmup_epoch: 2 + regularizer: + name: L2 + factor: 5.0e-05 +PostProcess: + name: DBPostProcess + thresh: 0.3 + box_thresh: 0.6 + max_candidates: 1000 + unclip_ratio: 1.5 +Metric: + name: DetMetric + main_indicator: hmean +Train: + dataset: + name: SimpleDataSet + data_dir: ../chinese/ + label_file_list: + - ../chinese/zhongce_training_fix_1.6k.txt + - ../chinese/label_train_all_f4_part2.txt + - ../chinese/label_train_all_f4_part4.txt + - ../chinese/label_train_all_f4_part5_no_born-digit_ifly5000.txt + - ../chinese/synch_en_ch_largeword_ppocr_2W.txt + - ../chinese/pubtabnet_ppocr_5W_new.txt + - ../chinese/label_synth_sheet_1W.txt + ratio_list: + - 1 + - 0.1 + - 0.1 + - 0.4 + - 0.05 + - 0.04 + - 0.05 + transforms: + - DecodeImage: + img_mode: BGR + channel_first: false + - DetLabelEncode: null + - IaaAugment: + augmenter_args: + - type: Fliplr + args: + p: 0.5 + - type: Affine + args: + rotate: + - -10 + - 10 + - type: Resize + args: + size: + - 0.5 + - 3 + - EastRandomCropData: + size: + - 960 + - 960 + max_tries: 50 + keep_ratio: true + - MakeBorderMap: + shrink_ratio: 0.4 + thresh_min: 0.3 + thresh_max: 0.7 + - MakeShrinkMap: + shrink_ratio: 0.4 + min_text_size: 8 + - NormalizeImage: + scale: 1./255. + mean: + - 0.485 + - 0.456 + - 0.406 + std: + - 0.229 + - 0.224 + - 0.225 + order: hwc + - ToCHWImage: null + - KeepKeys: + keep_keys: + - image + - threshold_map + - threshold_mask + - shrink_map + - shrink_mask + loader: + shuffle: true + drop_last: false + batch_size_per_card: 8 + num_workers: 4 +Eval: + dataset: + name: SimpleDataSet + data_dir: ../chinese/test_set/ + label_file_list: + - ../chinese/test_set/Label_refine_v8.txt + transforms: + - DecodeImage: + img_mode: BGR + channel_first: false + - DetLabelEncode: null + - DetResizeForTest: null + - NormalizeImage: + scale: 1./255. + mean: + - 0.485 + - 0.456 + - 0.406 + std: + - 0.229 + - 0.224 + - 0.225 + order: hwc + - ToCHWImage: null + - KeepKeys: + keep_keys: + - image + - shape + - polys + - ignore_tags + loader: + shuffle: false + drop_last: false + batch_size_per_card: 1 + num_workers: 2 diff --git a/configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_student.yml b/configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_student.yml new file mode 100644 index 00000000..d081c26a --- /dev/null +++ b/configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_student.yml @@ -0,0 +1,178 @@ +Global: + debug: false + use_gpu: true + epoch_num: 500 + log_smooth_window: 20 + print_batch_step: 10 + save_model_dir: ./output/ch_PP-OCR_V3/ + save_epoch_step: 100 + eval_batch_step: + - 0 + - 400 + load_static_weights: false + cal_metric_during_train: false + pretrained_model: null + checkpoints: null + save_inference_dir: null + use_visualdl: false + infer_img: doc/imgs_en/img_10.jpg + save_res_path: ./checkpoints/det_db/predicts_db.txt + distributed: true + student_pretrained_model: ./pretrain_model/MobileNetV3_large_x0_5_pretrained + teacher_pretrained_model: ./inference/ch_ppocr_resnet50/student +Architecture: + model_type: det + algorithm: DB + Transform: + Backbone: + name: MobileNetV3 + scale: 0.5 + model_name: large + disable_se: True + Neck: + name: FEPAN + out_channels: 96 + Head: + name: DBHead + k: 50 + kernel_size: [7, 2, 2] + +Loss: + name: DBLoss + balance_loss: true + main_loss_type: DiceLoss + alpha: 5 + beta: 10 + ohem_ratio: 3 +Optimizer: + name: Adam + beta1: 0.9 + beta2: 0.999 + lr: + name: Cosine + learning_rate: 0.001 + warmup_epoch: 2 + regularizer: + name: L2 + factor: 5.0e-05 +PostProcess: + name: DBPostProcess + thresh: 0.3 + box_thresh: 0.6 + max_candidates: 1000 + unclip_ratio: 1.5 +Metric: + name: DetMetric + main_indicator: hmean +Train: + dataset: + name: SimpleDataSet + data_dir: ../chinese/ + label_file_list: + - ../chinese/zhongce_training_fix_1.6k.txt + - ../chinese/label_train_all_f4_part2.txt + - ../chinese/label_train_all_f4_part4.txt + - ../chinese/label_train_all_f4_part5_no_born-digit_ifly5000.txt + - ../chinese/synch_en_ch_largeword_ppocr_2W.txt + - ../chinese/pubtabnet_ppocr_5W_new.txt + - ../chinese/label_synth_sheet_1W.txt + ratio_list: + - 1 + - 0.1 + - 0.1 + - 0.4 + - 0.05 + - 0.04 + - 0.05 + transforms: + - DecodeImage: + img_mode: BGR + channel_first: false + - DetLabelEncode: null + - IaaAugment: + augmenter_args: + - type: Fliplr + args: + p: 0.5 + - type: Affine + args: + rotate: + - -10 + - 10 + - type: Resize + args: + size: + - 0.5 + - 3 + - EastRandomCropData: + size: + - 960 + - 960 + max_tries: 50 + keep_ratio: true + - MakeBorderMap: + shrink_ratio: 0.4 + thresh_min: 0.3 + thresh_max: 0.7 + - MakeShrinkMap: + shrink_ratio: 0.4 + min_text_size: 8 + - NormalizeImage: + scale: 1./255. + mean: + - 0.485 + - 0.456 + - 0.406 + std: + - 0.229 + - 0.224 + - 0.225 + order: hwc + - ToCHWImage: null + - KeepKeys: + keep_keys: + - image + - threshold_map + - threshold_mask + - shrink_map + - shrink_mask + loader: + shuffle: true + drop_last: false + batch_size_per_card: 8 + num_workers: 4 +Eval: + dataset: + name: SimpleDataSet + data_dir: ../chinese/test_set/ + label_file_list: + - ../chinese/test_set/Label_refine_v8.txt + transforms: + - DecodeImage: + img_mode: BGR + channel_first: false + - DetLabelEncode: null + - DetResizeForTest: null + - NormalizeImage: + scale: 1./255. + mean: + - 0.485 + - 0.456 + - 0.406 + std: + - 0.229 + - 0.224 + - 0.225 + order: hwc + - ToCHWImage: null + - KeepKeys: + keep_keys: + - image + - shape + - polys + - ignore_tags + loader: + shuffle: false + drop_last: false + batch_size_per_card: 1 + num_workers: 2 -- GitLab