Aihub 금융 10% test score | from the scratch (Aihub 금융 train) | pretrained
(no transfer learned) | transfer learned
(Aihub 금융 train) |
bbox precision | 0.9662 | 0.0853 | 0.9678 |
bbox recall | 0.9797 | 0.6240 | 0.9824 |
bbox hmean | 0.9729 | 0.1502 | 0.9750 |
from the scratch
pretrained
transfer learned
•
10% from the scratch
nohup tools/dist_train.sh \
configs/textdet/dbnet/dbnet_resnet18_fpnc_20e_aihubfinance10of100.py \
2 > nohup.out &
Bash
복사
train
cp work_dirs/dbnet_resnet18_fpnc_20e_aihubfinance10of100/epoch_20.pth \
pretrained/dbnet_resnet18_fpnc_20e_aihubfinance10of100_sparkling-cloud-104.pth
nohup tools/dist_test.sh \
configs/textdet/dbnet/dbnet_resnet18_fpnc_20e_aihubfinance10of100.py \
pretrained/dbnet_resnet18_fpnc_20e_aihubfinance10of100_sparkling-cloud-104.pth \
2 > nohup.out &
Bash
복사
eval
11/28 19:12:01 - mmengine - INFO - Epoch(test) [3250/3250] AihubFinance10of100/icdar/precision: 0.9538 AihubFinance10of100/icdar/recall: 0.9713 AihubFinance10of100/icdar/hmean: 0.9625 AihubFinance100of100/icdar/precision: 0.9253 AihubFinance100of100/icdar/recall: 0.9582 AihubFinance100of100/icdar/hmean: 0.9415 IC15/icdar/precision: 0.0000 IC15/icdar/recall: 0.0000 IC15/icdar/hmean: 0.0000
Bash
복사
eval res
python3 -m mmocr.ocr \
--det-config configs/textdet/dbnet/dbnet_resnet18_fpnc_20e_aihubfinance10of100.py \
--det-ckpt pretrained/dbnet_resnet18_fpnc_20e_aihubfinance10of100_sparkling-cloud-104.pth \
data/det/aihub_finance/part_10of100/imgs/IMG_OCR_6_F_00964.png \
--img-out-dir work_dirs/dbnet_resnet18_fpnc_20e_aihubfinance10of100 \
--pred-out-file work_dirs/dbnet_resnet18_fpnc_20e_aihubfinance10of100/output.pkl \
--device cpu
Bash
복사
io
python3 -m work_dirs_utils.pkl2json \
work_dirs/dbnet_resnet18_fpnc_20e_aihubfinance10of100/output.pkl \
work_dirs/dbnet_resnet18_fpnc_20e_aihubfinance10of100/output.json
Bash
복사
jsonify