微调训练后,原来能识别的数据现在无法识别了。 #16404
Unanswered
MrWangChong
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
🔎 Search before asking
🐛 Bug (问题描述)
我没有修改PP-OCRv5_server_rec.yml文件,训练命令:
python3 -m paddle.distributed.launch --gpus '0,1' tools/train.py -c configs/rec/PP-OCRv5/PP-OCRv5_server_rec.yml -o Global.pretrained_model=./PP-OCRv5_server_rec_pretrained.pdparams
训练结果如下:
我取的第43轮训练的结果
从测试集看效果不错,使用训练的图片测试效果也不错。
但是使用新数据来测试效果很差,原有能识别的数据,现在也识别不了了,请问一下怎么做可以避免这种情况呢?需要调PP-OCRv5_server_rec.yml里面的参数吗
🏃♂️ Environment (运行环境)
Ubuntu 22.04.5 LTS (GNU/Linux 6.8.0-65-generic x86_64)
Python 3.10.12
paddleocr 3.2.0
paddlepaddle-gpu 3.1.1
paddlex 3.2.0
🌰 Minimal Reproducible Example (最小可复现问题的Demo)
from paddleocr import TableRecognitionPipelineV2
pipeline = TableRecognitionPipelineV2(
text_recognition_model_name="PP-OCRv5_server_rec",
text_recognition_model_dir="../../PP-OCRv5_server_rec_infer/"
)
import datetime
print("识别1开始时间"+str(datetime.datetime.now()))
output = pipeline.predict("./IMG_20250901_101548.jpg")
print("识别1完成时间"+str(datetime.datetime.now()))
for res in output:
res.print() ## 打印预测的结构化输出
res.save_to_img("tabocr1")
res.save_to_json("tabocr1")
Beta Was this translation helpful? Give feedback.
All reactions