What's the results of the given trained weights for ICDAR2015?

Question

What's the results of the given trained weights for ICDAR2015?

Closed this issue 5 years ago · 6 comments

Following the description, I download the trained weights, then test and evaluate on the detetion datasets of IC15. The results seems strange as follow:
precision: 0.966, recall: 0.744, hmean: 0.841
Is the result correct? Thanks for your reply.

Answer 1 · 2019-11-08T09:25:37.000Z

Hi!
Do I need an Australian insitution's account to download the weights from https://cloudstor.aarnet.edu.au/ ?
Thank you!

Answer 2 · 2019-11-08T13:49:38.000Z

@sisrfeng I just run the bash file and don't create any account.

Answer 3 · 2019-11-09T05:44:42.000Z

thank you!

Answer 4 · 2019-11-11T12:07:21.000Z

You can download the weights directly, without any account.

Answer 5 · 2019-11-11T12:08:49.000Z

@ZhouEnjoysL Note that the codes and models were provided for reimplementing E2E results on ICDAR 2015. For detection results, please directly use the outputs of the detection branch.

Answer 6 · 2020-02-21T09:19:38.000Z

Hello, have you solved this problem? I also want to test the performance of charnet and use it to run ICDAR2015 and Total_text, but I am a beginner and ca n’t do it. Can you teach me? Thank you.