FOTS.PyTorch
FOTS.PyTorch copied to clipboard
jiangxiluning
Code state
Hi, great repo !
Judging from the code the implementation is not complete, for example the Recognizer class is still not implemented while being used here. In this issue response you say that you are training the model, have you succeed in implementing the model ?
It would be great to have such info !
Thank you for your appreciation, I will speak out if the project fails or sucesses.
@jiangxiluning @victoresque What are your thoughts about AdvancedEAST? https://github.com/huoyijie/AdvancedEAST
It's a nice improvement of east, it directly predicts two ends of a text proposal avoiding predicting any intermediate pixel, which decreases the chance of breaking the long text proposal.
@jiangxiluning, hi luning, the http://rrc.cvc.uab.es/ is broken now, do you have an ICDAR2017 MLT dataset？ can you share it to me? sincerely thanks, my email is [email protected]
@jiangxiluning, very very thanks. when you finish the uploading, can you send the link to me by the email, I think the data link can not directly to be showed on public.
Hi Luning, thanks for the great work. May I know if you mind sending me the dataset link also? [email protected]
Thank you so much!
@jiangxiluning I might need ICDAR 2013, 2015 and 2017MLT. May I know if you happen to have any one or all of them? Sorry for the trouble!
@draculayuan http://rrc.cvc.uab.es/ is working, you can download the datasets there.
@MrOCR Thanks for replying! I tried using that cite, but it seems must register first before downloading their data. However, I tried using three different email to register but none of them received the activation link, so I still cant access to that data. May I know if you know of any ways to work around that issue? Thanks!
@jiangxiluning don't mind him. just focus on finishing the FOTS code. If he uses Gmail or Outlook it should work.
I tried 2 different gmails and 1 school outlook email two weeks back but none of that works. Got the data from CSDN website, still thanks for the suggestions. All the best for reimplementing the paper!
@MrOCR It seems FOTS doesn't have any mechanism for improving the performance of long text lines detection.
@jiangxiluning Though, can it detect single words on a long text-line? Thus, detecting only the words even if the text-line is long?
