Thank you for your impressive work on RoboFusion! I have read the paper and found the idea of integrating VFMs with 3D detection very inspiring.
I am currently trying to reproduce your results. However, I noticed that the repository currently provides instructions to manually modify existing frameworks (OpenPCDet/MMDetection3D) based on the .md files (e.g., Transfusion+SAM.md, FocalsConv+SAM.md).
Since manual modification is error-prone and might lead to version incompatibility issues, could you kindly provide a complete, integrated codebase (e.g., a forked version of OpenPCDet/MMDet3D with your changes already applied)?
Having a ready-to-run repository would significantly help the community to understand the exact model architecture and reproduce the performance reported in the paper.
Thanks for your help!
Thank you for your impressive work on RoboFusion! I have read the paper and found the idea of integrating VFMs with 3D detection very inspiring.
I am currently trying to reproduce your results. However, I noticed that the repository currently provides instructions to manually modify existing frameworks (OpenPCDet/MMDetection3D) based on the .md files (e.g., Transfusion+SAM.md, FocalsConv+SAM.md).
Since manual modification is error-prone and might lead to version incompatibility issues, could you kindly provide a complete, integrated codebase (e.g., a forked version of OpenPCDet/MMDet3D with your changes already applied)?
Having a ready-to-run repository would significantly help the community to understand the exact model architecture and reproduce the performance reported in the paper.
Thanks for your help!