aqibsaeed / place-recognition-using-autoencoders-and-nn Goto Github PK

Hi,
Thanks for your awesome implementation that makes our work easier.
Having run this code based on UJIIndoorLoc database, however, the original code cannot reach over 90% testing accuracy. I think it blames some settings in this program which is not definitely consistent with the detailed settings mentioned in the paper:

• Jointly scaling the raw material
• Using -110 to represent the lack of signal rather than 100 which is used in database.

These can help to achieve the expected performance according to paper.

i have read you paper and code,and get a lot,but i have a question:how to save or print the predicted result of the test data,that is the predicted label.because your code just give the final accuracy,no the predicted result.
thank you!!!!waiting for your response!