megvii-research PETR issues

megvii-research / PETR

[ECCV2022] PETR: Position Embedding Transformation for Multi-View 3D Object Detection & [ICCV2023] PETRv2: A Unified Framework for 3D Perception from Multi-Camera Images

Other

878 stars 132 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

论文中的3D position embedding similarity 可视化

#168 sidiangongyuan opened 1 month ago
0
请问代码不支持 torch2.2以上的版本吗？

#167 mmglove opened 2 months ago
0
How can I train PETRv2 in "current frame" only mode?

#166 Jeong-Bin opened 3 months ago
0
depth issue

#165 noahstudyer opened 4 months ago
0
不太理解为什么bin_size 使用两个不同的坐标轴进行计算

#164 EveningLin opened 4 months ago
0
Train Time

#163 ShenZheng2000 opened 4 months ago
0
Should I use 'Detr3Dlane_flatten200' for PETRv2_BEVseg?

#162 Jeong-Bin closed 2 months ago
1
PETRv1 stuck at high loss

#161 tungyen opened 5 months ago
0
Drivable area segmentation?

#160 Deltaidiots opened 7 months ago
0
RuntimeError: Given groups=1, weight of size [128, 256, 3, 3], expected input[1, 1024, 50, 50] to have 256 channels, but got 1024 channels instead

#159 loserhou opened 7 months ago
0
position_embeding

#158 L1bertad opened 7 months ago
0
cuda11.2 and pytorch1.9.1

#157 xudh1991 opened 8 months ago
1
How to export to ONNX ?

#156 AlexIswarya opened 9 months ago
2
请问，如果使用4个鱼眼相机作为输入，需要修改哪些地方？

#155 bigliuyq opened 9 months ago
2
How to train Waymo Dataset?

#154 becauseofAI opened 10 months ago
0
why no encoder like detr2

#153 xudh1991 opened 10 months ago
0
robustness analysis

#152 wofeifeileni opened 10 months ago
2
about 4*4 transfom matrix

#151 charlesliangcai closed 10 months ago
0
您好，请问一下点云数据在整个dataload的过程中起到什么作用呢？

#150 XYX1998 closed 10 months ago
0
Could you provide the config and checkpoint of PETRV2-BEVseg of Resnet101?

#149 aoyanl opened 11 months ago
0
Regarding the use of PETRv2 for semantic segmentation

#148 LisaShen0509 opened 11 months ago
4
Any results or suggestions if i want to use it on ScanNet?

#147 SerCharles opened 1 year ago
0
3D position_embeding in frustrume

#146 BingyuZhou opened 1 year ago
0
How to convert to onnx and then deploy?

#145 qiuhaining opened 1 year ago
0
"2f" is only use to debug since "30f" loads very slowly. It can sample from the "30f".

#144 Vendulamrdka95 opened 1 year ago
1
Hi,

#143 Vendulamrdka95 opened 1 year ago
5
> > > Hi,

#142 Vendulamrdka95 opened 1 year ago
1
grad_norm:nan

#141 losetiger opened 1 year ago
3
Ask for the code of visualization of attention map

#140 mezhou closed 10 months ago
5
Single-Camera Images

#139 becauseofAI opened 1 year ago
0
开源3DLane

#138 daishiqiang123 opened 1 year ago
1
Does Petrv2_head predict detection and segmentation at the same time?

#137 hagianga21 opened 1 year ago
0
FP16 training problem

#136 Legend94rz opened 1 year ago
0
Application of PETRv2 on OpenLanev2 lane topology task

#135 cyty98 closed 7 months ago
3
Ask for the weights of flash attention model

#134 2401ch opened 1 year ago
0
petr seg head only support batch size 1 ?

#133 FDInSky opened 1 year ago
1
Parameter at index 667 with name pts_lane_head.transformer_lane.decoder.layers.5.norms.2.weight has been marked as ready twice. This means that multiple autograd engine hooks have fired for this particular parameter during this iteration.

#132 FDInSky closed 1 year ago
0
Lane head?

#131 becauseofAI opened 1 year ago
0
Bev SegHead

#130 github-lpw opened 1 year ago
0
will open source lane's code ? Thanks

#129 FDInSky opened 1 year ago
2
Evaluate the released models but get very low NDS (mAP is ok)

#128 PeizeSun closed 1 year ago
2
Question about augmentation

#127 jianingwangind opened 1 year ago
8
i can't load mmdet3d_nuscenes_30f_infos_train.pkl

#126 dsw-jlu-rgzn opened 1 year ago
4
Will add streampetr into this repo?

#125 FDInSky opened 1 year ago
1
Package conflicts when installing mmdet3d (error: networkx 2.2 is installed but networkx>=2.8 is required by {'scikit-image'})

#124 da-qing-wa opened 1 year ago
8
About denoising

#123 jianingwangind closed 1 year ago
1
Why does the network uses only the first output of FPN ?

#122 LoickCh closed 1 year ago
3
Normalized 3D coordinates or un-normalized coordinates?

#121 sithu31296 closed 1 year ago
8
为什么用这么多转置矩阵？

#120 shaoqb opened 1 year ago
0
How to visualize the BEV segmentation results?

#119 linlion0311 opened 1 year ago
3