openocr-python 0.0.9__py3-none-any.whl → 0.1.0.dev0__py3-none-any.whl

openocr/__init__.py

@@ -8,4 +8,38 @@ __dir__ = os.path.dirname(os.path.abspath(__file__))
 sys.path.append(__dir__)
 sys.path.insert(0, os.path.abspath(os.path.join(__dir__, '..')))
 
-from tools.infer_e2e import OpenOCR, OpenDetector, OpenRecognizer
+# from .tools.infer_e2e import OpenOCRE2E, OpenDetector, OpenRecognizer
+# from .tools.infer_unirec_onnx import UniRecONNX
+# from .tools.infer_doc_onnx import OpenDocONNX
+from .openocr import OpenOCR, main
+
+__version__ = '0.1.0.dev'
+
+# Lazy import for demo interfaces to avoid initialization on import
+def launch_openocr_demo(*args, **kwargs):
+    """Launch Gradio OCR demo"""
+    from .demo_gradio import launch_demo
+    return launch_demo(*args, **kwargs)
+
+def launch_unirec_demo(*args, **kwargs):
+    """Launch UniRec demo"""
+    from .demo_unirec import launch_demo
+    return launch_demo(*args, **kwargs)
+
+def launch_opendoc_demo(*args, **kwargs):
+    """Launch OpenDoc demo"""
+    from .demo_opendoc import launch_demo
+    return launch_demo(*args, **kwargs)
+
+__all__ = [
+    'OpenOCRE2E',
+    'OpenDetector',
+    'OpenRecognizer',
+    'UniRecONNX',
+    'OpenDocONNX',
+    'OpenOCR',
+    'main',
+    'launch_openocr_demo',
+    'launch_unirec_demo',
+    'launch_opendoc_demo',
+]

openocr/configs/dataset/rec/evaluation.yaml

@@ -0,0 +1,41 @@
+root: ../evaluation
+task: str
+download_links:
+  # IC15_1811
+  - https://drive.usercontent.google.com/download?id=1eGY0kXNV1qVxeUpoGzs-ioUO-ky7msH6&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1BWv7aLoLAT7avY326gXP3GJF48UZpuBC&authuser=0&confirm=t
+  # SVT
+  - https://drive.usercontent.google.com/download?id=1ecEZ4cJ7dIbTCZRltE0s5KzUotQWagH-&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1OygBP7i9R-3Pwi6WodCcW31J8CUMugOJ&authuser=0&confirm=t
+  # IIIT5k
+  - https://drive.usercontent.google.com/download?id=1PJ9_IvIGZTS5hHdGLnpKuYKZcCO8jE0E&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=10P3MixSBt1v8k8_6aFfziC33Z5IlM6Uf&authuser=0&confirm=t
+  # IC13_857
+  - https://drive.usercontent.google.com/download?id=1-wMHOFBXJaOaY-UD00nDn6qw2s_8R4Vd&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1J1QCFtOFxFKiLJIgTqZ6eRo9Y5QGqHpA&authuser=0&confirm=t
+  # SVTP
+  - https://drive.usercontent.google.com/download?id=1kckwfZkdaHG8k_FW5IIJKUaYZkF21Hza&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1x61lm_ea7lvIdxNPMG-jy-5W0MxtdH0N&authuser=0&confirm=t
+  # CUTE80
+  - https://drive.usercontent.google.com/download?id=1Zv_91c81tinLy5Je89HPr-5wUSnqXKIB&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1OuJ6QoJ9AlyNHIM9j2WedAPxTnac7kyY&authuser=0&confirm=t
+filenames:
+  # IC15_1811
+  - ../evaluation/IC15_1811/data.mdb
+  - ../evaluation/IC15_1811/lock.mdb
+  # SVT
+  - ../evaluation/SVT/data.mdb
+  - ../evaluation/SVT/lock.mdb
+  # IIIT5k
+  - ../evaluation/IIIT5k/data.mdb
+  - ../evaluation/IIIT5k/lock.mdb
+  # IC13_857
+  - ../evaluation/IC13_857/data.mdb
+  - ../evaluation/IC13_857/lock.mdb
+  # SVTP
+  - ../evaluation/SVTP/data.mdb
+  - ../evaluation/SVTP/lock.mdb
+  # CUTE80
+  - ../evaluation/CUTE80/data.mdb
+  - ../evaluation/CUTE80/lock.mdb
+check_validity: true

openocr/configs/dataset/rec/ltb.yaml

@@ -0,0 +1,9 @@
+root: ../ltb
+task: str
+download_links:
+  - https://drive.usercontent.google.com/download?id=16AEA1YGTsyVB44uEjKi4ZUV1snjCYBr4&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1xU4OStrOaI23bPG4flWAPWn2YrQe2bmY&authuser=0&confirm=t
+filenames:
+  - ../ltb/data.mdb
+  - ../ltb/lock.mdb
+check_validity: true

openocr/configs/dataset/rec/mjsynth.yaml

@@ -0,0 +1,11 @@
+root: ../synth
+task: str
+download_links:
+  - https://drive.usercontent.google.com/download?id=1FIoplSFZ-BKQoRDHDXsVMKa844e-K8PD&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1eckTvaeRtlTZvbO2orrVz-cIuIk6i87K&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1PBXTf-2PnmEvJBsqzJqxxRwzhAZGTiMG&authuser=0&confirm=t
+filenames:
+  - ../synth/MJ_train.zip
+  - ../synth/MJ_val.zip
+  - ../synth/MJ_test.zip
+check_validity: true

openocr/configs/dataset/rec/openvino.yaml

@@ -0,0 +1,25 @@
+root: ../OpenVINO
+task: str
+download_links:
+  # train_1
+  - https://drive.usercontent.google.com/download?id=1q23QAIRTyG0t-bBm4aAwRwiqB6VUfphw&authuser=0&confirm=
+  # train_2
+  - https://drive.usercontent.google.com/download?id=1AtbaJljM68cbZqi5lcM92d9VkQUCbSqI&authuser=0&confirm=
+  # train_5
+  - https://drive.usercontent.google.com/download?id=1dejstYnJ8_sESuO_uvwi__jT1B8gPxf3&authuser=0&confirm=t
+  # train_f
+  - https://drive.usercontent.google.com/download?id=1C4akchTc7-yi1OS_sJ3KP693UKcnecke&authuser=0&confirm=t
+  # validation
+  - https://drive.usercontent.google.com/download?id=17TRzSQhuK_juAxAv3KmX0y13pQP2cz6R&authuser=0&confirm=t
+filenames:
+  # train_1
+  - ../OpenVINO/train_1.zip
+  # train_2
+  - ../OpenVINO/train_2.zip
+  # train_5
+  - ../OpenVINO/train_5.zip
+  # train_f
+  - ../OpenVINO/train_f.zip
+  # validation
+  - ../OpenVINO/validation.zip
+check_validity: true

openocr/configs/dataset/rec/ost.yaml

@@ -0,0 +1,17 @@
+root: ../OST
+task: str
+download_links:
+  # OST heavy
+  - https://drive.usercontent.google.com/download?id=1RGpIFbD_SRlrzZFBoVF_LGvetNx1-5pg&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1Th4MfDf44k0EBpIqCLqVoGRu6G-FP1hq&authuser=0&confirm=t
+  # OST weak
+  - https://drive.usercontent.google.com/download?id=1z5CTDJucUnvALG12Q4UXk1DDKJDd8WJn&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1V17TTkX3sjpV7v0km_F2SDCK0tL3k_ls&authuser=0&confirm=t
+filenames:
+  # OST heavy
+  - ../OST/heavy/data.mdb
+  - ../OST/heavy/lock.mdb
+  # OST weak
+  - ../OST/weak/data.mdb
+  - ../OST/weak/lock.mdb
+check_validity: true

openocr/configs/dataset/rec/synthtext.yaml

@@ -0,0 +1,7 @@
+root: ../synth
+task: str
+download_links:
+  - https://drive.usercontent.google.com/download?id=1T-enqkq6_l2HqrsV3da_h0oJ7CUKu_oc&authuser=0&confirm=t
+filenames:
+  - ../synth/ST.zip
+check_validity: true

openocr/configs/dataset/rec/test.yaml

@@ -0,0 +1,77 @@
+root: ../test
+task: str
+download_links:
+  # IC13_857
+  - https://drive.usercontent.google.com/download?id=1PZSCbe6_DI8MlCqCRWXGT2PP92_frIXq&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1qkN7NDg0zUHxUiZHAeEatDTqlsgpFWp3&authuser=0&confirm=t
+  # IC15_2077
+  - https://drive.usercontent.google.com/download?id=1dFkY3DNbr-Mepn3TWBiA9COEJ63fGFcp&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1UvVwLNZ3tS1YdTBa8MulPzjeVezKaDro&authuser=0&confirm=t
+  # SVTP
+  - https://drive.usercontent.google.com/download?id=1aofeerilxJ7J3S7QxuCEXbmXTpz8Xshx&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1rJ1KoO4K_VUxEAUN_bMgBGzK8_JZAAno&authuser=0&confirm=t
+  # IIIT5k
+  - https://drive.usercontent.google.com/download?id=1XFO2M1Kbgwv3-iTNTmhQXAEjNmKYOeoT&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1stwK2hFsyaV7HHsEG9EYgnUQebNb2_nG&authuser=0&confirm=t
+  # COCOv1.4
+  - https://drive.usercontent.google.com/download?id=1Se2QSGS19xx7Gfy-SUdX9mlAOr2eYsfA&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1xvekFi389QfkH7yS0JIVV0QzjhUspjDv&authuser=0&confirm=t
+  # IC15_1811
+  - https://drive.usercontent.google.com/download?id=1pHsw8wrThD9EGEE6AusQLZozefSj4iyR&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1TXZ1qHuKAksaAlvd3qMv4IHKnN-IJW9a&authuser=0&confirm=t
+  # Uber
+  - https://drive.usercontent.google.com/download?id=1L2j6BZeLTGQ1FIl8HB_D3AFiWLltGV5r&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=12DUj28yzLWxFO_gfMfSjTkRujYD5MNEE&authuser=0&confirm=t
+  # IC13_1095
+  - https://drive.usercontent.google.com/download?id=1fu8onMt3Z6fDLNAiHcm-sQ2qCXduE-FU&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1OQAZtLj8U2Cl4L0ErGFsz6vGIVTTWasD&authuser=0&confirm=t
+  # IC13_1015
+  - https://drive.usercontent.google.com/download?id=1mbsfuvWB282HYfn9tbqcj1nUDkLXcSNB&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1QGogU_hV-oN7iY2POutdD2LDcmK6plnV&authuser=0&confirm=t
+  # ArT
+  - https://drive.usercontent.google.com/download?id=1-53knSy-uTSngCG7wyBngVyTuTCmdnWl&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=172EsSaf7BVaB1ORtohi-Jc_8SuUKZGGf&authuser=0&confirm=t
+  # SVT
+  - https://drive.usercontent.google.com/download?id=1p7aVUr9Yr7c4X4YUBvk2-YP28rraHjn9&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1ALmhvSleZ0yf-lcdbQPP3M9Zc3oqnXij&authuser=0&confirm=t
+  # CUTE80
+  - https://drive.usercontent.google.com/download?id=1Ujr4axHKnu54P2rIGUhkjdM6XlhDYrI_&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1DvZi9L3MqjO2zRUyCg3YvP4qMAt2bsme&authuser=0&confirm=t
+filenames:
+  # IC13_857
+  - ../test/IC13_857/data.mdb
+  - ../test/IC13_857/lock.mdb
+  # IC15_2077
+  - ../test/IC15_2077/data.mdb
+  - ../test/IC15_2077/lock.mdb
+  # SVTP
+  - ../test/SVTP/data.mdb
+  - ../test/SVTP/lock.mdb
+  # IIIT5k
+  - ../test/IIIT5k/data.mdb
+  - ../test/IIIT5k/lock.mdb
+  # COCOv1.4
+  - ../test/COCOv1.4/data.mdb
+  - ../test/COCOv1.4/lock.mdb
+  # IC15_1811
+  - ../test/IC15_1811/data.mdb
+  - ../test/IC15_1811/lock.mdb
+  # Uber
+  - ../test/Uber/data.mdb
+  - ../test/Uber/lock.mdb
+  # IC13_1095
+  - ../test/IC13_1095/data.mdb
+  - ../test/IC13_1095/lock.mdb
+  # IC13_1015
+  - ../test/IC13_1015/data.mdb
+  - ../test/IC13_1015/lock.mdb
+  # ArT
+  - ../test/ArT/data.mdb
+  - ../test/ArT/lock.mdb
+  # SVT
+  - ../test/SVT/data.mdb
+  - ../test/SVT/lock.mdb
+  # CUTE80
+  - ../test/CUTE80/data.mdb
+  - ../test/CUTE80/lock.mdb
+check_validity: true

openocr/configs/dataset/rec/textocr.yaml

@@ -0,0 +1,13 @@
+root: ../TextOCR
+task: str
+download_links:
+  # train
+  - https://drive.usercontent.google.com/download?id=1jVjJFno4pnsU0Cp_kn4MIXQrChmELy92&authuser=0&confirm=
+  # val
+  - https://drive.usercontent.google.com/download?id=1ubIRu01MXIek6OvInu-XjaIbw6277-vw&authuser=0&confirm=t
+filenames:
+  # train
+  - ../TextOCR/train.zip
+  # val
+  - ../TextOCR/val.zip
+check_validity: true

openocr/configs/dataset/rec/textocr_horizontal.yaml

@@ -0,0 +1,13 @@
+root: ../TextOCR_horizontal
+task: str
+download_links:
+  # train
+  - https://drive.usercontent.google.com/download?id=1sWH6J11xbjQb8SH7fdG_8mIKVI81ZQy5&authuser=0&confirm=
+  # val
+  - https://drive.usercontent.google.com/download?id=1gIE-AU2o-5hvg288-bjphO6UkI5AEQ2d&authuser=0&confirm=t
+filenames:
+  # train
+  - ../TextOCR_horizontal/train.zip
+  # val
+  - ../TextOCR_horizontal/val.zip
+check_validity: true

openocr/configs/dataset/rec/union14m_b.yaml

@@ -0,0 +1,47 @@
+root: ../u14m
+task: str
+download_links:
+  # artistic
+  - https://drive.usercontent.google.com/download?id=1Je2DTuFHnkXDI99yDnm9Anl5naWaCQwd&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1xtT_Q0juBJUIvAG55qBxoVNNTECd2usZ&authuser=0&confirm=t
+  # contextless
+  - https://drive.usercontent.google.com/download?id=1_0OzyzWhZOmGrHkayFTVrzhrQrNRDRPR&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1PPgC42y3xoM9bR0HQFbDYbcT3PzMdD_y&authuser=0&confirm=t
+  # salient
+  - https://drive.usercontent.google.com/download?id=1tHLMYBmTqRnxvFOTT3dfLfQiundqFWfd&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=13NQgpAtCK0kh9M5E2pAUmKKEp6Qu5Xwj&authuser=0&confirm=t
+  # multi_words
+  - https://drive.usercontent.google.com/download?id=1IlnDKX3V_Vp9gsDGFB0xoqsVLH1vtxUI&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1mFFjC7C0CwevvkwFU9YeVbZBdps_3Qpb&authuser=0&confirm=t
+  # curve
+  - https://drive.usercontent.google.com/download?id=1MxhMd85cmhUtI2lmtXhZQuFk7lav0_fw&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1N03g-4e-kJG2mRvlM0c5TrwWAkd-iG-Q&authuser=0&confirm=t
+  # general
+  - https://drive.usercontent.google.com/download?id=1Oqt7OaycP466NWoDmoJ3FqS8YP3YRgvu&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1K0MrX5eYNt8IIGFHXCwg0_oI5OF5PPFO&authuser=0&confirm=t
+  # multi_oriented
+  - https://drive.usercontent.google.com/download?id=1TKZFcZPVk0ThqfF-AGhJk_OCLg0ykKbv&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1PAoLMUWuR7O2-7XRoKkNzQcSiznErQzD&authuser=0&confirm=t
+filenames:
+  # artistic
+  - ../u14m/artistic/data.mdb
+  - ../u14m/artistic/lock.mdb
+  # contextless
+  - ../u14m/contextless/data.mdb
+  - ../u14m/contextless/lock.mdb
+  # salient
+  - ../u14m/salient/data.mdb
+  - ../u14m/salient/lock.mdb
+  # multi_words
+  - ../u14m/multi_words/data.mdb
+  - ../u14m/multi_words/lock.mdb
+  # curve
+  - ../u14m/curve/data.mdb
+  - ../u14m/curve/lock.mdb
+  # general
+  - ../u14m/general/data.mdb
+  - ../u14m/general/lock.mdb
+  # multi_oriented
+  - ../u14m/multi_oriented/data.mdb
+  - ../u14m/multi_oriented/lock.mdb
+check_validity: true

openocr/configs/dataset/rec/union14m_l_filtered.yaml

@@ -0,0 +1,35 @@
+root: ../Union14M-L-LMDB-Filtered
+task: str
+download_links:
+  # train_challenging
+  - https://drive.usercontent.google.com/download?id=1etwzBgGHjsFsb0sygsaRnKbanW2PMe07&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1ly6FJfPjItwGlVQ-ifTrzzM3rVu3Ezhr&authuser=0&confirm=t
+  # train_easy
+  - https://drive.usercontent.google.com/download?id=1_zeNluTnywIaa5h3PN-Ah9tKyByypot7&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1caYLeQHDidXgVBDi9IWXbO1gg__DYq9a&authuser=0&confirm=t
+  # train_hard
+  - https://drive.usercontent.google.com/download?id=1eP6s2xyYPZX9gykvWA4VSOc3Fqul_UB_&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1-ZlCvocX8P5uVRclUXp_5DNGLDzd16EO&authuser=0&confirm=t
+  # train_medium
+  - https://drive.usercontent.google.com/download?id=1s_CoaLNJEr-UxHYiqZ5jOcliMCFiRUUy&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1Wpj6WVpZ5Ily77kVwfQ18CiZBzkgmEnF&authuser=0&confirm=t
+  # train_normal
+  - https://drive.usercontent.google.com/download?id=1jPt44arlAswl9cXZjzmVcdpptdTPpJ3I&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1Rfc5kE03AzOUv7B_eYcBhUV8KMQ2MZ1m&authuser=0&confirm=t
+filenames:
+  # train_challenging
+  - ../Union14M-L-LMDB-Filtered/train_challenging/data.mdb
+  - ../Union14M-L-LMDB-Filtered/train_challenging/lock.mdb
+  # train_easy
+  - ../Union14M-L-LMDB-Filtered/train_easy/data.mdb
+  - ../Union14M-L-LMDB-Filtered/train_easy/lock.mdb
+  # train_hard
+  - ../Union14M-L-LMDB-Filtered/train_hard/data.mdb
+  - ../Union14M-L-LMDB-Filtered/train_hard/lock.mdb
+  # train_medium
+  - ../Union14M-L-LMDB-Filtered/train_medium/data.mdb
+  - ../Union14M-L-LMDB-Filtered/train_medium/lock.mdb
+  # train_normal
+  - ../Union14M-L-LMDB-Filtered/train_normal/data.mdb
+  - ../Union14M-L-LMDB-Filtered/train_normal/lock.mdb
+check_validity: true

openocr/configs/rec/cmer/cmer.yml

@@ -0,0 +1,127 @@
+Global:
+  device: gpu
+  epoch_num: 1
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/cmer
+  save_epoch_step: [0, 1]
+  save_batch_step: [30000, 10000]
+  eval_epoch_step: [200, 1]
+  eval_batch_step: [100000000, 2000]
+  total_iter_steps: 1000000
+  cal_metric_during_train: False
+  # pretrained_model: /path/to/pretrained/model.pth
+  checkpoints:
+  resume_from_iter: False
+  use_tensorboard: false
+  infer_img: ./cmer_test_image
+  max_text_length: &max_text_length 1024
+  use_space_char: &use_space_char True
+  save_res_path: ./output/cmer/infer_results.txt
+  use_amp: True
+  use_ema: False
+  use_transformers: True
+  grad_clip_val: 1.0
+
+Optimizer:
+  name: AdamW
+  lr: 3.0e-5
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+  fused: True
+
+LRScheduler:
+  name: WarmupCosineLR
+  warmup_steps: 5000
+  eta_min: 3.0e-7
+
+
+Architecture:
+  model_type: rec
+  algorithm: CMER
+  in_channels: 3
+  Transform:
+  Encoder:
+  Decoder:
+    out_channels: -1
+  vision_config:
+      num_layers: 8
+      num_heads: 16
+      hidden_dim: 1024
+      down_sample_ratio: 32
+  decoder_config:
+    attention_bias: false
+    attention_dropout: 0.0
+    bos_token_id: 2
+    decoder_layers: 8
+    eos_token_id: 3
+    head_dim: 16
+    hidden_act: "gelu"
+    hidden_size: 1024
+    initializer_range: 0.02
+    intermediate_size: 3072
+    layer_types: null
+    max_position_embeddings: 4096
+    max_window_layers: 28
+    num_attention_heads: 16
+    num_key_value_heads: 16
+    pad_token_id: 1
+    rms_norm_eps: 1.0e-06
+    rope_scaling: null
+    rope_theta: 100000.0
+    sliding_window: 4096
+    tie_word_embeddings: false
+    use_cache: true
+    use_sliding_window: true
+    vocab_size: 23948
+
+Loss:
+  name: CMERLoss
+
+PostProcess:
+  name: CMERLabelDecode
+
+Train:
+  dataset:
+    name: CMERWebDataSet
+    data_dir: "/path/to/MER-17M"
+    max_text_length: 1024
+    processor: CMERProcessor
+    processor_source: openrec.preprocess.cmer_label_encode
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - CMERProcessor:
+
+  loader:
+    batch_size_per_card: 16
+    drop_last: True
+    shuffle: False
+    num_workers: 8
+    pin_memory: True
+
+
+# no use
+Eval:
+  dataset:
+    name: CMERWebDataSet
+    data_dir: "/path/to/MER-17M"
+    processor: CMERProcessor
+    processor_source: openrec.preprocess.cmer_label_encode
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - CMERProcessor:
+
+  loader:
+    batch_size_per_card: 8
+    drop_last: False
+    shuffle: False
+    num_workers: 8
+    pin_memory: False
+
+
+Metric:
+  name: "CMERMetric"
+  args:
+    main_indicator: bleu

openocr/configs/rec/mdiff4str/svtrv2_mdiffdecoder_base.yml

@@ -0,0 +1,152 @@
+Global:
+  device: gpu
+  epoch_num: 40
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtrv2_mdiffdecoder_base/
+  save_epoch_step: [30, 1]
+  # evaluation is run every 2000 iterations
+  eval_batch_step: [0, 500]
+  eval_epoch_step: [0, 1]
+  cal_metric_during_train: False
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: &character_dict_path ./tools/utils/EN_symbol_dict.txt # 96en
+  # ./tools/utils/ppocr_keys_v1.txt  # ch
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_mdiffdecoder_base.txt
+  use_amp: True
+  grad_clip_val: 20.0
+
+Optimizer:
+  name: AdamW
+  lr: 0.0005 # for 8gpus bs128/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: MDiff4STR
+  in_channels: 3
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: False
+  Decoder:
+    name: MDiffDecoder
+    num_decoder_layers: 6
+    nhead: 6
+    max_len: *max_text_length
+    parallel_decoding: False
+    autoregressive_decoding: False
+    low_confidence_decoding: False
+    random_mask_decoding: False
+    semi_autoregressive_decoding: False
+    cloze_mask_decoding: False
+    sampler_step: 3
+    sample_k: &sample_k 3
+    temperature: 1.0
+
+Loss:
+  name: MDiffLoss
+
+PostProcess:
+  name: ARLabelDecode
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: false
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - MDiffLabelEncode: # Class handling label
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+          sample_num: *sample_k
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'reflect_ids', 'noisy_batch', 'masked_indices', 'p_mask', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 4
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: False
+    data_dir_list: [
+      '../evaluation/CUTE80',
+      '../evaluation/IC13_857',
+      '../evaluation/IC15_1811',
+      '../evaluation/IIIT5k_3000',
+      '../evaluation/SVT',
+      '../evaluation/SVTP',
+      ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ARLabelEncode: # Class handling label
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: *bs
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: False
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: *bs
+    max_ratio: *max_ratio
+    num_workers: 4