This is the official repository for ''ER3D: An Efficient Real-time 3D Object Detection Framework for Autonomous Driving''
Due to the double-blind principle, this repository including the video is published as an anonymous resource.
[Video]
Download the official KITTI 3D object detection dataset and organize the downloaded files as follows:
ER3D
├── data
│ ├── kitti
│ │ ├── ImageSets
│ │ │ ├── val.txt ...
│ │ ├── training
│ │ │ ├── calib /000000.txt ...
│ │ │ ├── image_2 /000000.png ...
│ │ │ ├── image_3 /000000.png ...
We provide the optimized Dynamic Library of Semi-Global Matching (SGM) of our implementation. To run it properly, installing the environments with the same version as ours would be necessary.
- GPU: TITAN RTX
- opencv=3.4.12
- cuda=10.2
- cudnn=7.6.5
conda install pytorch=1.10.1 torchvision=0.11.2 cudatoolkit=10.2
pip install spconv-cu102
cd utils/iou3d_nms
python setup.py develop
python test.py