Правки requirements и README.
Some checks are pending
Build/Publish Develop Docs / deploy (push) Waiting to run

This commit is contained in:
2025-07-03 12:20:32 +03:00
parent 63e910b982
commit 0d9e36bc92
4 changed files with 30 additions and 14 deletions

View File

@@ -1,9 +1,24 @@
```bash ```bash
python -m venv paddle python -m venv paddle
source paddle/bin/activate source paddle/bin/activate
pip install -r deploy/requirements pip install -r requirements.txt
mkdir pretrained_models/ mkdir pretrained_models/
cd pretrained_models/ cd pretrained_models/
wget https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-OCRv5_server_rec_pretrained.pdparams wget https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-OCRv5_server_rec_pretrained.pdparams
cd .. cd ..
mkdir train_data/
cd train_data/
```
Далее положите датасет custom-ocr-dataset.zip в папку train_data и далее выполните:
```bash
unzip custom-ocr-dataset.zip
rm custom-ocr-dataset.zip
cd ..
```
Запустите процесс валидации:
```bash
python tools/eval.py -c configs/rec_custom.yml
``` ```

View File

@@ -1,8 +1,8 @@
Global: Global:
use_gpu: true use_gpu: false
character_dict_path: /kaggle/input/custom-ocr-dataset/dict.txt character_dict_path: train_data/dict.txt
save_model_dir: ./output/ save_model_dir: ./output/
pretrained_model: /kaggle/working/PaddleOCR/pretrained_models/PP-OCRv5_server_rec_pretrained.pdparams pretrained_model: pretrained_models/PP-OCRv5_server_rec_pretrained.pdparams
Architecture: Architecture:
model_type: rec model_type: rec
@@ -38,9 +38,9 @@ Metric:
Train: Train:
dataset: dataset:
name: SimpleDataSet name: SimpleDataSet
data_dir: /kaggle/input/custom-ocr-dataset data_dir: train_data/
label_file_list: label_file_list:
- /kaggle/input/custom-ocr-dataset/train.txt - train_data/train.txt
loader: loader:
batch_size_per_card: 16 # Уменьшите для экономии памяти batch_size_per_card: 16 # Уменьшите для экономии памяти
num_workers: 2 num_workers: 2
@@ -48,9 +48,9 @@ Train:
Eval: Eval:
dataset: dataset:
name: SimpleDataSet name: SimpleDataSet
data_dir: /kaggle/input/custom-ocr-dataset data_dir: train_data/
label_file_list: label_file_list:
- /kaggle/input/custom-ocr-dataset/val.txt - train_data/val.txt
transforms: transforms:
- DecodeImage: - DecodeImage:
img_mode: BGR img_mode: BGR

View File

@@ -1,6 +0,0 @@
# paddlepaddle-gpu
paddlepaddle
paddleocr
imgaug
lmdb
rapidfuzz

View File

@@ -15,3 +15,10 @@ albumentations
# to be compatible with albumentations # to be compatible with albumentations
albucore albucore
packaging packaging
# paddlepaddle-gpu
paddlepaddle
paddleocr
imgaug
lmdb
rapidfuzz
setuptools