[1_1] MMOCR 에서 제공되는 임의의 detection 모델을 샘플 데이터(MINI-ICDAR2015)로 학습한다.

mini ICDAR visualization

Quick Run - MMOCR 1.0.0rc3 documentation

Please refer to here for a quick inference run. A detailed description of MMOCR's inference interface can be found here Note In addition to using our well-provided pre-trained models, you can also train models on your own datasets.

https://mmocr.readthedocs.io/en/dev-1.x/get_started/quick_run.html

python3 -m tools.analysis_tools.browse_dataset \
configs/textdet/dbnet/dbnet_resnet18_fpnc_1200e_icdar2015_mini.py \
--output-dir ./artifacts/browse_dataset/mini_icdar2015
Python
복사

학습

mini ICDAR 2015 학습

Quick Run - MMOCR 1.0.0rc3 documentation

Please refer to here for a quick inference run. A detailed description of MMOCR's inference interface can be found here Note In addition to using our well-provided pre-trained models, you can also train models on your own datasets.

https://mmocr.readthedocs.io/en/dev-1.x/get_started/quick_run.html

# python3 -m tools.train {config_file_path}
python3 -m tools.train \
configs/textdet/dbnet/dbnet_resnet18_fpnc_1200e_icdar2015_mini.py
Bash
복사

2-gpu

Training and Testing - MMOCR 1.0.0rc3 documentation

To meet diverse requirements, MMOCR supports training and testing models on various devices, including PCs, work stations, computation clusters, etc. tools/train.py provides the basic training service. MMOCR recommends using GPUs for model training and testing, but it still enables CPU-Only training and testing.

https://mmocr.readthedocs.io/en/dev-1.x/user_guides/train_test.html

# tools/dist_train.sh {config_file_path} {n_gpu}
tools/dist_train.sh \
configs/textdet/dbnet/dbnet_resnet18_fpnc_1200e_icdar2015_mini.py \
2
Bash
복사

평가

F1 Score 산출

mini ICDAR 2015 가중치 로드 및 추론

Quick Run - MMOCR 1.0.0rc3 documentation

Please refer to here for a quick inference run. A detailed description of MMOCR's inference interface can be found here Note In addition to using our well-provided pre-trained models, you can also train models on your own datasets.

https://mmocr.readthedocs.io/en/dev-1.x/get_started/quick_run.html

# python3 -m tools.test {config_file_path} {pth_path}
python3 -m tools.test \
configs/textdet/dbnet/dbnet_resnet18_fpnc_1200e_icdar2015_mini.py \
work_dirs/dbnet_resnet18_fpnc_1200e_icdar2015_mini/epoch_400.pth
Bash
복사

2-gpu

Training and Testing - MMOCR 1.0.0rc3 documentation

To meet diverse requirements, MMOCR supports training and testing models on various devices, including PCs, work stations, computation clusters, etc. tools/train.py provides the basic training service. MMOCR recommends using GPUs for model training and testing, but it still enables CPU-Only training and testing.

https://mmocr.readthedocs.io/en/dev-1.x/user_guides/train_test.html

# tools/dist_test.sh {config_file_path} {pth_path} {n_gpu}
tools/dist_test.sh \
configs/textdet/dbnet/dbnet_resnet18_fpnc_1200e_icdar2015_mini.py \
work_dirs/dbnet_resnet18_fpnc_1200e_icdar2015_mini/epoch_400.pth \
2
Bash
복사