rasd3 / 3d-cvf Goto Github PK
View Code? Open in Web Editor NEW[ECCV 2020] This is the official implementation of 3D-CVF: Generating Joint Camera and LiDAR Features Using Cross-View Spatial Feature Fusion for 3D Object Detection
[ECCV 2020] This is the official implementation of 3D-CVF: Generating Joint Camera and LiDAR Features Using Cross-View Spatial Feature Fusion for 3D Object Detection
hi, when I run 'from spconv.utils import rbbox_iou, rbbox_iou_loss' in box_np_ops.py, it occurs an error: ImportError: cannot import name 'rbbox_iou_loss'. However, when I ask the author about rrbox_iou_loss, he said there isn't any rbbox_iou_loss in spconv. So how should I modify the code?
Traceback (most recent call last):
File "train.py", line 19, in
from second.builder import target_assigner_builder, voxel_builder
File "/home/codes/deeplearn/3d/3D-CVF/second/builder/target_assigner_builder.py", line 3, in
from second.core.target_assigner import TargetAssigner
File "/home/codes/deeplearn/3d/3D-CVF/second/core/target_assigner.py", line 1, in
from second.core import box_np_ops
File "/home/codes/deeplearn/3d/3D-CVF/second/core/box_np_ops.py", line 5, in
from second.core.non_max_suppression.nms_gpu import rotate_iou_gpu_eval
File "/home/codes/deeplearn/3d/3D-CVF/second/core/non_max_suppression/init.py", line 2, in
from second.core.non_max_suppression.nms_gpu import (nms_gpu, rotate_iou_gpu,
File "/home/codes/deeplearn/3d/3D-CVF/second/core/non_max_suppression/nms_gpu.py", line 11, in
import iou3d_nms_cuda
ModuleNotFoundError: No module named 'iou3d_nms_cuda'
I use the envirment "docker pull yckimm/second:second_v1.5", and get this error.
could you fix it
Hi @Jaekyumkim ,thank you for your outstanding work. I am trying to run the second stage of the code. In the class VoxelNet (in voxelnet_second.py), self.rpn
is initialized as rpn_class_dic["RPN_SECOND_FUSION"]
, and self.second_rpn
is initialized as rpn_class_dict["RPN_FUSION"]
. Is it reversed?
Also the part second_preds_dict = self.second_rpn(bev_crops_output, concat_crops_output)
does't match the code in rpn.py. The number of input parameters is not even enough. Is there some code missing in rpn.py?
Hello, may I know when would the code be published? I'm so interested about your great work.
May I know when would the code be published?
hi, I found that in your code your train the model with the step number but not max_num_epoch, so what's the meaning of the max_num_epoch?
When I minimize the batch_size from 12 to 4, do I have to set steps for three times?
Has anyone run the second training stage successfully? Seems like the PointNet encoding part after RoI pooling is missing.
There is white arrow on the second/Rotated_ROIAlign file, and no file in it.
Could you help me?
The term "auto-calibrated projection" essentially refers to a "deformable transformer," correct?
idx_offset is not defined in train.py.
what is the mean of idx_offset?
how can i solve this problem?
thanks!
Can you provide more details, such as pre-training models
Traceback (most recent call last):
File "pytorch/train.py", line 16, in
from second.builder import target_assigner_builder, voxel_builder
File "/home/lzh/3dcode/fuse/3D-CVF/second/builder/target_assigner_builder.py", line 3, in
from second.core.target_assigner import TargetAssigner
File "/home/lzh/3dcode/fuse/3D-CVF/second/core/target_assigner.py", line 1, in
from second.core import box_np_ops
File "/home/lzh/3dcode/fuse/3D-CVF/second/core/box_np_ops.py", line 7, in
from spconv.utils import rbbox_iou, rbbox_iou_loss
ImportError: cannot import name 'rbbox_iou_loss' from 'spconv.utils' (/home/lzh/anaconda3/envs/i3dcvf/lib/python3.7/site-packages/spconv/utils/init.py)
thank you very much **
Hi,
Thanks for sharing the code of your method. I am trying to reproduce your results with your code but I am missing the pre-trained weight files you are using to initialize your networks: pre_weight/FPN18_retinanet_968.pth
and pre_weight/first_stage/fusion_split/voxelnet-35210.tckpt
.
Is there a way to download those files somewhere or pre-train the networks in a similar way ?
could you share with us more details about data augmentation, which are used in paper? thank you for your sharing!
Has anyone run this project? Can you get the results in the paper ?
Hi, Thanks for sharing the excellent works. I wonder what voxel size did you use for the model on the nuScenes test set?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.