For reproducing the code effect

[lmz-sense]

Thank you very much for your contribution to the field of deep estimation and for providing me with excellent research ideas! But when debugging the effectiveness of your code, I encountered a problem: I followed the debugging steps you provided step by step to test. Without adding any pre trained models, my accuracy is far from reaching your experimental results. The experimental results are as follows:

If the pre training model you provided was loaded, the training results would get worse and worse. Later, I looked at the pre training weight file you provided and tested it, and found that the results seemed to be the experimental results of your paper. I would like to know if you added pre training weights during the experiment. If so, I hope you can provide them because they are very interested in your research and have great reference value!

[fengyi233]

Thanks for your attention! Some bugs have been found and I will fix them as soon as possible.

[lmz-sense]

Thanks for your attention! Some bugs have been found and I will fix them as soon as possible.

Thank you for your reply. Looking forward to your prompt update. I am particularly interested in your work!

[wangjiyuan9]

Have you ever fixed the bug?

[lmz-sense]

Hello author, I am very interested in your work. Since I left you a comment last time, I have researched and tried your experimental content multiple times. Could you please provide me with the pre trained weights you used during the experiment? If you do not use pre trained weights, it seems that the original effect of the experiment cannot be achieved. I hope you can provide it. This is very important for my future research. If possible, thank you in advance!

[wangjiyuan9]

Hello author, I am very interested in your work. Since I left you a comment last time, I have researched and tried your experimental content multiple times. Could you please provide me with the pre trained weights you used during the experiment? If you do not use pre trained weights, it seems that the original effect of the experiment cannot be achieved. I hope you can provide it. This is very important for my future research. If possible, thank you in advance!

I am curious, can we reproduce the performance in the paper using the training Eigen set?

[lmz-sense]

Hello author, I am very interested in your work. Since I left you a comment last time, I have researched and tried your experimental content multiple times. Could you please provide me with the pre trained weights you used during the experiment? If you do not use pre trained weights, it seems that the original effect of the experiment cannot be achieved. I hope you can provide it. This is very important for my future research. If possible, thank you in advance!

I am curious, can we reproduce the performance in the paper using the training Eigen set?

At present, it seems that my attempt is not feasible. Without using pre trained weights, there is still a certain gap between the final effect and the one in the article. However, it seems that the author did not provide the pre trained weights.