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

megvii-research/PETR Issues

3D Position Encoder implementation question in code.
Updated a month ago2
Regarding the use of PETRv2 for semantic segmentation
Updated a month ago4
How to visualize the BEV segmentation results?
Updated a month ago3
Drivable area segmentation?
Updated a month ago
How to export to ONNX ?
Updated a month ago2
grad_norm:nan
Updated 2 months ago3
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
Updated 2 months ago
position_embeding
Updated 2 months ago
Application of PETRv2 on OpenLanev2 lane topology task
Closed 2 months ago3
请问，如果使用4个鱼眼相机作为输入，需要修改哪些地方？
Updated 2 months ago1
cuda11.2 and pytorch1.9.1
Updated 2 months ago1
Package conflicts when installing mmdet3d (error: networkx 2.2 is installed but networkx>=2.8 is required by {'scikit-image'})
Updated 3 months ago8
Ask for the code of visualization of attention map
Closed 4 months ago1
petr seg head only support batch size 1 ?
Updated 4 months ago1
How to train Waymo Dataset?
Updated 4 months ago
robustness analysis
Updated 4 months ago1
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？
Closed 4 months ago
why no encoder like detr2
Updated 4 months ago
about 4*4 transfom matrix
Closed 5 months ago
Could you provide the config and checkpoint of PETRV2-BEVseg of Resnet101?
Updated 6 months ago
Any results or suggestions if i want to use it on ScanNet?
Updated 6 months ago
3D position_embeding in frustrume
Updated 8 months ago
开源3DLane
Updated 8 months ago1
How to convert to onnx and then deploy?
Updated 8 months ago
Hi,
Updated 8 months ago5
"2f" is only use to debug since "30f" loads very slowly. It can sample from the "30f".
Updated 8 months ago1
> > > Hi,
Updated 8 months ago1
Single-Camera Images
Updated 10 months ago
will open source lane's code ? Thanks
Updated 10 months ago2
Does Petrv2_head predict detection and segmentation at the same time?
Updated 10 months ago
FP16 training problem
Updated 10 months ago
Ask for the weights of flash attention model
Updated 10 months ago
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.
Closed a year ago
Lane head?
Updated a year ago
Bev SegHead
Updated a year ago
Question about augmentation
Updated a year ago8
where is the PETRv2_3DLane？
Updated a year ago3
Evaluate the released models but get very low NDS (mAP is ok)
Closed a year ago2
i can't load mmdet3d_nuscenes_30f_infos_train.pkl
Updated a year ago3
About denoising
Closed a year ago1
Will add streampetr into this repo?
Updated a year ago1
where is the temporal fusion code part in petr_v2?
Updated a year ago6
Normalized 3D coordinates or un-normalized coordinates?
Closed a year ago8
Why does the network uses only the first output of FPN ?
Closed a year ago
为什么用这么多转置矩阵？
Updated a year ago
Incorrect 3D IoU calculated by function LiDARInstance3DBoxes.overlaps
Closed a year ago2
数据集的results['sweeps']字段下面没有camera相关的keys。
Updated a year ago4
cannot pickle 'dict_keys' object
Closed a year ago4
augmented NuScene as inputs to petrv2_BEVseg.py - inference output 3d boxes shifted and scaled
Updated a year ago6
nuScenes test 榜单使用val集没
Updated a year ago1