The official code release of MT-SSD, a point-based single stage 3D detection model, is now available. Our code is based on OpenPCDet. To install it, please follow the command below.
Requirements: Linux (Ubuntu 18.04/20.04/21.04)
Python 3.9.0
PyTorch 1.7.1
CUDA 9.0 or higher (PyTorch 1.3+ needs CUDA 9.2+)
git clone https://github.com/qifeng22/MT-SSD.git && cd MT-SSD
pip install spconv-cuxxx ### replace cuxxx with your cuda version,such as cu102
python setup.py develop
For OpenPCDet, please refer to here for the Dataset Preparation.
python train.py --cfg_file cfgs/kitti_models/MT-SSD.yaml --batch_size 8 --epoch 80 ## kitti
python train.py --cfg_file cfgs/waymo_models/MT-SSD.yaml --batch_size 12 --epoch 85 ## waymo single GPU
sh scripts/dist_train.sh 2 --cfg_file cfgs/waymo_models/MT-SSD.yaml --epoch 85
## waymo multiple GPUs, you can change 2 to the number what you want.
For MT-SSD on waymo, we release the code in another repository, please refer to MT-SSD(waymo).
python test.py --cfg_file cfgs/kitti_models/IA-SSD.yaml --batch_size ${BATCH_SIZE} --ckpt ${PTH_FILE}
## PTH_FILE your_fold/MT-SSD/output/kitti_models/IA-SSD/default/ckpt/checkpoint_epoch_80.pth
python test.py --cfg_file cfgs/waymo_models/MT-SSD.yaml --batch_size ${BATCH_SIZE} --ckpt ${PTH_FILE}
Quantitative results of different approaches on KITTI dataset (test set):
Here is the ckpt .
Quantitative results of different approaches on KITTI dataset (val set):
Here is the ckpt
Here is the ckpt
