运行demo.py用时9秒，正常吗？

YCG09 / chinese_ocr

CTPN + DenseNet + CTC based end-to-end Chinese OCR implemented using tensorflow and keras

Apache License 2.0

2.76k stars 1.08k forks source link

Open gitmessage opened 5 years ago

gitmessage commented 5 years ago

ubuntu 16, gpu:Quadro P2000 运行python demo.py识别作者提供的图片。这个速度正常吗？多谢 Detection took 3.666s for 22 object proposals Mission complete, it took 9.334s

YCG09 commented 5 years ago

你试试批量识别下看看，第一次比较慢

wrsuifeng commented 5 years ago

Mac OS python2.7 运行demo也花了9s

gitmessage commented 5 years ago

demo.jpg复制了一份变成了demo2.jpg。两个文件，第二个文件的速度： Detection took 1.293s for 22 object proposals Mission complete, it took 1.700s 给作者点赞！

PeterBishop0 commented 5 years ago

请问我这个demo复制了多份执行速度还是很慢应该怎么办，平均6s

666888985 commented 4 years ago

@YCG09 ，作者你好，运行demo.py可以出现识别结果吗，为啥看到readme.md里面说的将测试图片放入test_images目录，检测结果会保存到test_result中，没有说识别结果？另外那个评估准确率，识别率，召回率的代码有吗？

666888985 commented 4 years ago

@YCG09 ，有没有完整的先检测再识别的demo代码？