About training details

[fred-lyu]

Hi, your work is amazing, and thanks for your contribution to the community. In our experiment, we were unable to achieve the accuracy mentioned in this paper. Therefore, I would like to ask how you trained on two datasets with inconsistent image sizes, as well as the specific training process. Thanks.

[fred-lyu]

This is the result of my 40 epoch of training on the nyu dataset. Can you tell me how to continue with the remaining 20 epoch of training on the kitti dataset. resolutions: 320 x 256 med | std | mean | & 0.461 & 0.251 & 0.470
abs_rel | sq_rel | rmse | rmse_log | a1 | a2 | a3 | & 0.139 & 0.008 & 0.039 & 0.179 & 0.827 & 0.953 & 0.989 \

[water221]

这是我在 nyu 数据集上进行 40 个 epoch 训练的结果。您能告诉我如何继续在 kitti 数据集上进行剩余 20 个 epoch 的训练吗？ 分辨率：320 x 256 中|标准|意思是| & 0.461 & 0.251 & 0.470 绝对相对值 | sq_rel |均方根误差 | rmse_log | a1 | a2 | a3| & 0.139 & 0.008 & 0.039 & 0.179 & 0.827 & 0.953 & 0.989 \

Hello, I have been reproducing this model recently, but I have an error in the process of running it, giving a prompt. There is no mpvit_base definition in the "./networks/mpvit.py" file. Here you can see that you have successfully run for 40 epoch. If it is convenient for you, can you send your mpvit.py file and ./splits/nyu/cpose folder? if possible, I would appreciate

[fred-lyu]

这是我训练用的mvpit.py文件，以及./splits/nyu文件夹。

This is the mvpit.py file I used for training, as well as/ Splits/nyu folder.

在 2024-04-07 18:31:18，"zhutianshui" @.***> 写道：

这是我在 nyu 数据集上进行 40 个 epoch 训练的结果。您能告诉我如何继续在 kitti 数据集上进行剩余 20 个 epoch 的训练吗？ 分辨率：320 x 256 中|标准|意思是| & 0.461 & 0.251 & 0.470 绝对相对值 | sq_rel |均方根误差 | rmse_log | a1 | a2 | a3| & 0.139 & 0.008 & 0.039 & 0.179 & 0.827 & 0.953 & 0.989 \

Hello, I have been reproducing this model recently, but I have an error in the process of running it, giving a prompt. There is no mpvit_base definition in the "./networks/mpvit.py" file. Here you can see that you have successfully run for 40 epoch. If it is convenient for you, can you send your mpvit.py file and ./splits/nyu/cpose folder? if possible, I would appreciate

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

从网易163邮箱发来的超大附件推荐客户端极速下载 mpvit.py (23.68K, 2024年4月27日 19:27 到期) 下载 nyu.zip (351.7M, 2024年4月27日 19:27 到期) 下载

[water221]

这是我训练用的mvpit.py文件，以及./splits/nyu文件夹。 This is the mvpit.py file I used for training, as well as/ Splits/nyu folder. 在 2024-04-07 18:31:18，"zhutianshui" @.> 写道： 这是我在 nyu 数据集上进行 40 个 epoch 训练的结果。您能告诉我如何继续在 kitti 数据集上进行剩余 20 个 epoch 的训练吗？ 分辨率：320 x 256 中|标准|意思是| & 0.461 & 0.251 & 0.470 绝对相对值 | sq_rel |均方根误差 | rmse_log | a1 | a2 | a3| & 0.139 & 0.008 & 0.039 & 0.179 & 0.827 & 0.953 & 0.989 \ Hello, I have been reproducing this model recently, but I have an error in the process of running it, giving a prompt. There is no mpvit_base definition in the "./networks/mpvit.py" file. Here you can see that you have successfully run for 40 epoch. If it is convenient for you, can you send your mpvit.py file and ./splits/nyu/cpose folder? if possible, I would appreciate — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.> 从网易163邮箱发来的超大附件推荐客户端极速下载 mpvit.py (23.68K, 2024年4月27日 19:27 到期) 下载 nyu.zip (351.7M, 2024年4月27日 19:27 到期) 下载

Sorry, there seems to be no link attached to your reply. I can't find a place to click the download link.

[fred-lyu]

You can find the files from this link.

https://web.ugreen.cloud/web/#/share/3bb43d3ac535407bb95094da19359f40 提取码：A5C3

在 2024-04-12 19:39:25，"zhutianshui" @.***> 写道：

这是我训练用的mvpit.py文件，以及./splits/nyu文件夹。 This is the mvpit.py file I used for training, as well as/ Splits/nyu folder. 在 2024-04-07 18:31:18，"zhutianshui" @.> 写道： 这是我在 nyu 数据集上进行 40 个 epoch 训练的结果。您能告诉我如何继续在 kitti 数据集上进行剩余 20 个 epoch 的训练吗？ 分辨率：320 x 256 中|标准|意思是| & 0.461 & 0.251 & 0.470 绝对相对值 | sq_rel |均方根误差 | rmse_log | a1 | a2 | a3| & 0.139 & 0.008 & 0.039 & 0.179 & 0.827 & 0.953 & 0.989 \ Hello, I have been reproducing this model recently, but I have an error in the process of running it, giving a prompt. There is no mpvit_base definition in the "./networks/mpvit.py" file. Here you can see that you have successfully run for 40 epoch. If it is convenient for you, can you send your mpvit.py file and ./splits/nyu/cpose folder? if possible, I would appreciate — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.> 从网易163邮箱发来的超大附件推荐客户端极速下载 mpvit.py (23.68K, 2024年4月27日 19:27 到期) 下载 nyu.zip (351.7M, 2024年4月27日 19:27 到期) 下载

Sorry, there seems to be no link attached to your reply. I can't find a place to click the download link.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

[water221]

You can find the files from this link. https://web.ugreen.cloud/web/#/share/3bb43d3ac535407bb95094da19359f40 提取码：A5C3 在 2024-04-12 19:39:25，"zhutianshui" @.> 写道： 这是我训练用的mvpit.py文件，以及./splits/nyu文件夹。 This is the mvpit.py file I used for training, as well as/ Splits/nyu folder. 在 2024-04-07 18:31:18，"zhutianshui" @.> 写道： 这是我在 nyu 数据集上进行 40 个 epoch 训练的结果。您能告诉我如何继续在 kitti 数据集上进行剩余 20 个 epoch 的训练吗？ 分辨率：320 x 256 中|标准|意思是| & 0.461 & 0.251 & 0.470 绝对相对值 | sq_rel |均方根误差 | rmse_log | a1 | a2 | a3| & 0.139 & 0.008 & 0.039 & 0.179 & 0.827 & 0.953 & 0.989 \ Hello, I have been reproducing this model recently, but I have an error in the process of running it, giving a prompt. There is no mpvit_base definition in the "./networks/mpvit.py" file. Here you can see that you have successfully run for 40 epoch. If it is convenient for you, can you send your mpvit.py file and ./splits/nyu/cpose folder? if possible, I would appreciate — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.> 从网易163邮箱发来的超大附件推荐客户端极速下载 mpvit.py (23.68K, 2024年4月27日 19:27 到期) 下载 nyu.zip (351.7M, 2024年4月27日 19:27 到期) 下载 Sorry, there seems to be no link attached to your reply. I can't find a place to click the download link. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.>

Thank you very much for sharing and patient answers. I think your answer helped me a lot. Also, if it is convenient, can you share how to use COLMAP for pretreatment? I tried many times and failed. I'd appreciate it if you could share.

[fred-lyu]

The following scripts and files contain the Python code I used to process the nyu2 dataset. The main instructions are as follows: Colmap_coarsePose. py: Directly call the colmap library interface colmap_with python() through Python, or call the system installed colmap interface colmap_with cpp through OS. I recommend using the colmap_with cpp method because it can use cuda acceleration. The Python version of the colmap library does not support cuda by default. CheckFile.py: I remember this was to check whether different sequences in the nuy2 dataset were successfully solved by colmap. Because some sequences cannot be successfully solved by colmap. Process_gt.py: Convert deep data to npz format. CPose. py: Convert colmap output format to a pose format supported by gasmono.

The attachment is the above explanatory file, as well as the script output file.

I hope it can be helpful to you.

https://web.ugreen.cloud/web/#/share/c8bfd1c7cdc74cf888346e50b3b6cf2c 提取码：44V6

在 2024-04-12 19:51:39，"zhutianshui" @.***> 写道：

You can find the files from this link. https://web.ugreen.cloud/web/#/share/3bb43d3ac535407bb95094da19359f40 提取码：A5C3 在 2024-04-12 19:39:25，"zhutianshui" @.> 写道： 这是我训练用的mvpit.py文件，以及./splits/nyu文件夹。 This is the mvpit.py file I used for training, as well as/ Splits/nyu folder. 在 2024-04-07 18:31:18，"zhutianshui" @.> 写道： 这是我在 nyu 数据集上进行 40 个 epoch 训练的结果。您能告诉我如何继续在 kitti 数据集上进行剩余 20 个 epoch 的训练吗？ 分辨率：320 x 256 中|标准|意思是| & 0.461 & 0.251 & 0.470 绝对相对值 | sq_rel |均方根误差 | rmse_log | a1 | a2 | a3| & 0.139 & 0.008 & 0.039 & 0.179 & 0.827 & 0.953 & 0.989 \ Hello, I have been reproducing this model recently, but I have an error in the process of running it, giving a prompt. There is no mpvit_base definition in the "./networks/mpvit.py" file. Here you can see that you have successfully run for 40 epoch. If it is convenient for you, can you send your mpvit.py file and ./splits/nyu/cpose folder? if possible, I would appreciate — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.> 从网易163邮箱发来的超大附件推荐客户端极速下载 mpvit.py (23.68K, 2024年4月27日 19:27 到期) 下载 nyu.zip (351.7M, 2024年4月27日 19:27 到期) 下载 Sorry, there seems to be no link attached to your reply. I can't find a place to click the download link. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.>

Thank you very much for your sharing and patient answers. I feel that your articles and answers are very helpful to me. In addition, if it is convenient, you can share how you use COLMAP for preprocessing. I have tried many times and failed. If you can share, I will be very grateful.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

[water221]

The following scripts and files contain the Python code I used to process the nyu2 dataset. The main instructions are as follows: Colmap_coarsePose. py: Directly call the colmap library interface colmap_with python() through Python, or call the system installed colmap interface colmap_with cpp through OS. I recommend using the colmap_with cpp method because it can use cuda acceleration. The Python version of the colmap library does not support cuda by default. CheckFile.py: I remember this was to check whether different sequences in the nuy2 dataset were successfully solved by colmap. Because some sequences cannot be successfully solved by colmap. Process_gt.py: Convert deep data to npz format. CPose. py: Convert colmap output format to a pose format supported by gasmono. The attachment is the above explanatory file, as well as the script output file. I hope it can be helpful to you. https://web.ugreen.cloud/web/#/share/c8bfd1c7cdc74cf888346e50b3b6cf2c 提取码：44V6 在 2024-04-12 19:51:39，"zhutianshui" @.> 写道： You can find the files from this link. https://web.ugreen.cloud/web/#/share/3bb43d3ac535407bb95094da19359f40 提取码：A5C3 在 2024-04-12 19:39:25，"zhutianshui" @.> 写道： 这是我训练用的mvpit.py文件，以及./splits/nyu文件夹。 This is the mvpit.py file I used for training, as well as/ Splits/nyu folder. 在 2024-04-07 18:31:18，"zhutianshui" @.> 写道： 这是我在 nyu 数据集上进行 40 个 epoch 训练的结果。您能告诉我如何继续在 kitti 数据集上进行剩余 20 个 epoch 的训练吗？ 分辨率：320 x 256 中|标准|意思是| & 0.461 & 0.251 & 0.470 绝对相对值 | sq_rel |均方根误差 | rmse_log | a1 | a2 | a3| & 0.139 & 0.008 & 0.039 & 0.179 & 0.827 & 0.953 & 0.989 \ Hello, I have been reproducing this model recently, but I have an error in the process of running it, giving a prompt. There is no mpvit_base definition in the "./networks/mpvit.py" file. Here you can see that you have successfully run for 40 epoch. If it is convenient for you, can you send your mpvit.py file and ./splits/nyu/cpose folder? if possible, I would appreciate — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.> 从网易163邮箱发来的超大附件推荐客户端极速下载 mpvit.py (23.68K, 2024年4月27日 19:27 到期) 下载 nyu.zip (351.7M, 2024年4月27日 19:27 到期) 下载 Sorry, there seems to be no link attached to your reply. I can't find a place to click the download link. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.> Thank you very much for your sharing and patient answers. I feel that your articles and answers are very helpful to me. In addition, if it is convenient, you can share how you use COLMAP for preprocessing. I have tried many times and failed. If you can share, I will be very grateful. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.>

Thank you again for your patience and answers. I will carefully learn from the documents you share. I hope everything goes well in your research, work and life.

[CUMTchenlv]

Thank you for your outstanding contribution. Have you reproduced the author's results? In addition, the COLOMP preprocessing method, mpvit.py file and ./splits/nyu/cpose folder you shared have expired. I would be grateful if you could share them again.

[fred-lyu]

@CUMTchenlv No problem! This link is permanently accessible for download: https://drive.google.com/drive/folders/1mjpv43eHyE7WRQn8VvCiPqLy0_vK52zg?usp=sharing

[CUMTchenlv]

Thank you for your generous sharing. I wish you success in your research and more achievements.