search

zju3dv / NeuralRecon

Code for "NeuralRecon: Real-Time Coherent 3D Reconstruction from Monocular Video", CVPR 2021 oral
https://zju3dv.github.io/neuralrecon/
Apache License 2.0
2.03k stars 294 forks source link

Why not use ScanNet provided mesh as ground truth? #117

Open fengziyue opened 1 year ago

fengziyue commented 1 year ago

Hi:

thank you for sharing this amazing work, I have a question:

since ScanNet already provided the GT mesh, why do we need to run the TSDF fusion again to prepare the GT TSDF data?

Thank you!

MarklearnML commented 1 year ago

because the model is to predict TSDF instead of predicting mesh， so we need to get TSDF to training.

fengziyue commented 1 year ago

Is it possible to directly generate GT TSDF fromGT Mesh?

On Thu, Oct 20, 2022 at 8:38 AM MarklearnML @.***> wrote:

because the model is to predict TSDF instead of predicting mesh， so we need to get TSDF to training.

— Reply to this email directly, view it on GitHub https://urldefense.com/v3/__https://github.com/zju3dv/NeuralRecon/issues/117*issuecomment-1285473126__;Iw!!PTd7Sdtyuw!VZVemECA2i0Fg6qEun99kmclANnnBnqIJumJlfKNkwk15L_Zt2HDD0-OT2qZI7V70_qByxTZnTB70aBJPNaqS-gpig$, or unsubscribe https://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/AFCA33SUHB6TJFNU73Z3FIDWEE4MZANCNFSM6AAAAAARITEVKY__;!!PTd7Sdtyuw!VZVemECA2i0Fg6qEun99kmclANnnBnqIJumJlfKNkwk15L_Zt2HDD0-OT2qZI7V70_qByxTZnTB70aBJPNYKZIwoiA$ . You are receiving this because you authored the thread.Message ID: @.***>

-- Thank you!

Sincerely, Ziyue Feng