search

vishwa91 / wire

wavelet implicit neural representations
MIT License
130 stars 15 forks source link

Question about the unmatched experiment result #8

Open zhendshen2 opened 1 year ago

zhendshen2 commented 1 year ago

Dear author：

Above all, It is really a nice work!

I am writing to ask a question about the unmatch experiment result. When I redo the experiment with no change of the parameter you use, my experiment result is not as strong as yours.

For example, when I redo the experiment of denoising, in the paper the result is 30.2, however, in my experiment, it is only 28.25. In other experiments, there are similar situations.

Could I know is there any other adjustment you do to the parameters to cause this difference?

Best wishes, Zhenda

vishwa91 commented 1 year ago

Hi Zhenda,

Thank you for the note. Could you change the image resizing to 1/2 instead of 1/4 as it is in the code and rerun the experiment? Please let me know if any other experiments have a mismatch.

Regards.

zhendshen2 commented 1 year ago

Dear author:

Thank you very much for your answer!

I am also having a question about the 3D reconstruction (the occupancy problems). With no change of the parameter, the rerun result is only 0.97. However, the result in the paper is 0.99. Could I also ask is there anything necessary to be adjusted?

Meanwhile, for the super-resolution task. With no change of the parameter, the rerun result is only 23.2, which is also lower than what it is in the paper. Could I also ask is there anything necessary to be adjusted?

Thank you again for your support!

Best wishes, Zhenda Shen

vishwa91 commented 1 year ago

I am currently traveling but will respond within a week's time!

Regarding 3D occupancy: Please use omega=20, sigma=40.

For SISR: Will upload a working code within a week.

Thank you.