install.md

Step-by-step installation instructions

a. Clone EchoFusion.

git clone https://github.com/tusen-ai/EchoFusion
cd EchoFusion
mkdir ckpts ###pretrain weights
mkdir data ###dataset

The requied ckpts/resnet50_bev.pth can be downloaded from HERE.

b. Create a conda virtual environment and activate it.

conda create -n echofusion python=3.7 -y
conda activate echofusion

c. Install PyTorch and torchvision following the official instructions.

conda install pytorch==1.9.1 torchvision==0.10.1 torchaudio==0.9.1 cudatoolkit=11.1 -c pytorch -c conda-forge
# Recommended torch>=1.9

d. Install mmcv-full.

pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/{cu_version}/{torch_version}/index.html

examples:

pip install mmcv-full==1.4.0 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.9.1/index.html

e. Install mmdet and mmseg.

pip install mmdet==2.19.0
pip install mmsegmentation==0.20.0

f. Install mmdet3d from source code.

git clone https://github.com/open-mmlab/mmdetection3d.git
cd mmdetection3d
git checkout v0.17.3 

Then use mmdet3d_supp/core/bbox/iou_calculators provided by EchoFusion repo to add support of let-iou.

rm -rf ./mmdet3d/core/bbox/iou_calculators
cp -r ../mmdet3d_supp/core/bbox/iou_calculators ./mmdet3d/core/bbox
rm -rf ./mmdet3d/core/bbox/__init__.py
cp ../mmdet3d_supp/core/bbox/__init__.py ./mmdet3d/core/bbox
pip install -v -e .
cd ..

g. Install RADIal from source code following the official instructions.

git clone https://github.com/valeoai/RADIal
cd RADIal/DBReader
pip install .
conda install -c intel intel-aikit-modin
pip install cupy-cuda111
pip3 install pkbar
pip3 install polarTransform
pip3 install --upgrade git+https://github.com/klintan/pypcd.git
cd ../..

h. Install modified Extensible-Object-Detection-Evaluator from provided code.

cd Extensible-Object-Detection-Evaluator
pip install treelib
pip install ipdb
pip install .
cd ..

i. Other dependencies.

pip install cplxmodule
pip install spconv-cu113
conda update ffmpeg
pip install opencv-python