search

LiewFeng / RayDN

[ECCV 2024] Ray Denoising (RayDN): Depth-aware Hard Negative Sampling for Multi-view 3D Object Detection
https://arxiv.org/abs/2402.03634
Other
78 stars 1 forks source link

Inquiry on real time camera inference #12

Closed Darkthor15 closed 1 month ago

Darkthor15 commented 1 month ago

Hey! Great work and congratulations on your work being accepted by ECCV 2024! I actually wanted to use this to run inference on my camera feed for object detection can you please guide me through all the changes or what else i need to do in order to run it?

LiewFeng commented 1 month ago

Hi, thank you for your interest in our work. RayDN takes several frames of muliti-view(i.e., 6-view) images as input. Besides, it also requires the transformation matirx of the camera. You may follow StreamPETR and nuScenes to prepare your data.

Darkthor15 commented 1 month ago

Hi Thank you for your reply and time! Do i need to always convert a dataset to nuscenes format? does it support any other? Also how can i just test it on 1 or 2 images?

LiewFeng commented 1 month ago

I haven't tried it. You may refer to link.

Darkthor15 commented 1 month ago

Thanks a lot! I will refer to it and let you know if i have something else!

seawei1993 commented 4 weeks ago

hi , Have you successfully implemented real time inference? If possible, can you share it?

LiewFeng commented 3 weeks ago

Maybe you can also refer to link.