search

FangShancheng / ABINet

Read Like Humans: Autonomous, Bidirectional and Iterative Language Modeling for Scene Text Recognition
Other
429 stars 72 forks source link

Failed to reproduce results in the paper #16

Closed mandal4 closed 3 years ago

mandal4 commented 3 years ago

Thanks for release your source code. Unfortunately, i got poor test result (about -4% from official results in the paper)

Is there any important training strategies to reproduce the results? Or, is there anyone who meets same issue..?

FangShancheng commented 3 years ago

Thanks for reporting this issue, can you please provide your training environments and logs, and the accuracy on each dataset? It is unlikely to have a big gap using this code and I'll try to find out the reason.

FangShancheng commented 3 years ago

@mandal4 Have you resolved this problem, or training details can be provided? Does the training images you use be cropped directly from original datasets, or after perspective transform?

mandal4 commented 3 years ago

I found that my training environments are different with official setting

FangShancheng commented 3 years ago

Can you give the differences about the training environments that cause poor performance, which would be helpful.