Real-time-Text-Detection-DBNet icon indicating copy to clipboard operation
Real-time-Text-Detection-DBNet copied to clipboard

Training model

Open WJinhai opened this issue 6 years ago • 6 comments

The trained model can be public? Is it consistent with the results in the paper?

WJinhai avatar Dec 09 '19 02:12 WJinhai

The sufflenetv2 model trianed on ic13 and15 has been released, but for some implementation details are different, the results are not as good as the results in the paper. I will continue to update this project with the reference to the official implementation.

SURFZJY avatar Dec 10 '19 02:12 SURFZJY

Does this only work on GPU?

WJinhai avatar Dec 16 '19 08:12 WJinhai

It can work on CPU, by setting the gpu_id=None and uncommenting torch.cuda.synchronize(self.device).

SURFZJY avatar Dec 16 '19 09:12 SURFZJY

This result looks not as good as in the paper

WJinhai avatar Dec 17 '19 06:12 WJinhai

yes, i am working hard to improve the performance. I will release a better version some days later.

SURFZJY avatar Dec 17 '19 08:12 SURFZJY

@SURFZJY Did the results got better? whats the result on icdar15

ghost avatar Mar 19 '20 16:03 ghost