ttaoretw / imgeonet Goto Github PK

Hi it's great work!

I am very interested in the supervision of the geometry shaping module. If am not wrong, the input of the geometry module is V (HxWxDxC). If will go through several Conv3D and Conv3D(T) layers and output the geometry shaping weight, which has the same size with V. I am wondering the detailed steps of getting the ground truth surface voxels.

Assume we have 20 images per scene, then the "RGB-D frames" here means converting the 20 depth images to 20 sparse point clouds, and if one voxel doesn't contain any point of the 20 point clouds, the ground truth value will be negative or 0, and if one voxel contain points from the 20 point clouds, the ground truth value will be positive or 1?

Another question is about "for each camera ray, we also consider locations neighboring surface voxels within margin as positive". what does "for each camera ray" means here? In this geometry shaping step, the multi-view features are already fused. I am confused about the steps of selecting the neighbors.

The third question is the size of the ground truth surface voxel. Is it a HxWxDxC tensor with values 0 and 1? Then you apply focal loss to it and the predicted weight to supervise the geometry shaping model?

Really looking forward to your kind reply. Thank you very much!

When the code will be released?

I can't wait to try your excellent work!

Best wishes~

Hi, it's interesting work! I have a question about the performance of ImvoxelNet on ARKitScenes dataset stated in the paper.
The mAP0.25 and mAP0.5 for ImvoxelNet on ARKitScenes dataset in ImGeoNet paper is very high:

But in NeRF-Det and CN-RMA paper, the mAP0.25 of ImvoxelNet on ARKitScenes is less than 30:


Could you please kindly share how you train ImvoxelNet and your model on ARKitScenes and share the code?
Thank you very much! Looking forward to your reply.