The measuring unit of MSE in utils.py

[guoyage]

Hello, I have the confusion about RMSE. First of all, the RMSE calculation of the Polar model in the code is performed in the polar coordinate system, which brings about the first problem: in the polar coordinate system, the dimensions of radius and angle are not the same, and the two dimensions are different. Is it reasonable to add the values directly? Secondly, I tried to convert the predicted trajectory from the polar coordinate system to the Cartesian coordinate system. At this time, I found that the RMSE in the Cartesian coordinate system would increase, which brought about the second confusion: the difference between Polar and Cartesian coordinates. Is the RMSE comparison fair at this time? Very much looking forward to your reply.

[Petrichor625]

Thank you for your interest in our work!

Firstly, we would like to clarify that our model predicts future trajectories in Cartesian coordinates. Therefore, there is no need to convert coordinate systems during training. RMSE calculation includes proper unit conversions (feet to meters), ensuring fair comparison. Please note that for efficiency, printed results during training are in feet, whereas testing data is converted to metre for consistency. Testing on the validation set will demonstrate the final results.

Additionally, the version currently uploaded may not be the latest. We kindly request your patience as we work to update with the latest content. Thank you again for your patience.

[guoyage]

Thank you for your reply, but in your article, polar coordinate transformation is one of the features of your model, why do you say that no polar coordinate system is used now?

---- Replied Message ---- | From | Haicheng @.> | | Date | 04/30/2024 22:00 | | To | @.> | | Cc | @.>@.> | | Subject | Re: [Petrichor625/BATraj-Behavior-aware-Model] The measuring unit of MSE in utils.py (Issue #8) |

Thank you for your interest in our work!

Firstly, we would like to clarify that our model predicts future trajectories in Cartesian coordinates. Therefore, there is no need to convert coordinate systems during training. RMSE calculation includes proper unit conversions (feet to meters), ensuring fair comparison. Please note that for efficiency, printed results during training are in feet, whereas testing data is converted to metre for consistency. Testing on the validation set will demonstrate the final results.

Additionally, the version currently uploaded may not be the latest. We kindly request your patience as we work to update with the latest content. Thank you again for your patience.

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

[Petrichor625]

I think there might be a misunderstanding, or maybe we haven't clarified the issue properly. It's important to emphasize that we do use polar coordinate transformation as one of the features in our model. However, during the final decoding stage, if I recall correctly, our output results are in Cartesian coordinates and undergo unit conversions accordingly. In other words, while the historical trajectories entered into the model are transformed into polar coordinates, the final output results are transformed back into Cartesian coordinate. （Please be patient with us as we update the version of the code！）

[guoyage]

I think there might be a misunderstanding, or maybe we haven't clarified the issue properly. It's important to emphasize that we do use polar coordinate transformation as one of the features in our model. However, during the final decoding stage, if I recall correctly, our output results are in Cartesian coordinates and undergo unit conversions accordingly. In other words, while the historical trajectories entered into the model are transformed into polar coordinates, the final output results are transformed back into Cartesian coordinate. （Please be patient with us as we update the version of the code！）

Thank you for your reply and look forward to your reply

[Petrichor625]

Thank you all for your interest in our paper, and I apologize for the delayed response.

At the moment, there is an issue with the version of the code currently available in this repository, and unfortunately, we’re unable to upload the latest corrected version right now. However, for those looking to replicate or improve upon the results, I recommend exploring our other project, HLTP (GitHub Link), which provides more effective methods and enhanced code that may be beneficial to your work.

If you have any questions or run into challenges with HLTP, please don’t hesitate to reach out! As I may not always receive notifications from GitHub or have the chance to check issues regularly, please feel free to contact me via email if you need a quicker response. I’ll do my best to assist you as soon as possible.

Thank you again for your patience, understanding, and support—it’s truly appreciated!