Skip to content

Latest commit

 

History

History
56 lines (40 loc) · 1.54 KB

Installation.md

File metadata and controls

56 lines (40 loc) · 1.54 KB

Installation

The Installation step is referenced from CenterNet.


Our experimental environment:

Ubuntu 16.04.1, Python 3.5.2, PyTorch 0.4.1, torchvision 0.2.1.

During training we use 8 NVIDIA TITAN XP with cuda 9.0.


  1. Create a new conda environment and activate the environment.

    conda create --name MOC python=3.5.2
conda activate MOC

  2. Install pytorch0.4.1:

    conda install pytorch=0.4.1 torchvision -c pytorch

    Disable cudnn batch normalization(follow CenterNet).

    # PYTORCH=/path/to/pytorch # usually ~/anaconda3/envs/MOC/lib/python3.5.2/site-packages/
# for pytorch v0.4.1
sed -i "1254s/torch\.backends\.cudnn\.enabled/False/g" ${PYTORCH}/torch/nn/functional.py

    For other pytorch version, you can manually open torch/nn/functional.py and find the line with torch.batch_norm and replace the torch.backends.cudnn.enabled with False.

  3. Clone this repo (${MOC_ROOT} is the path to clone):

    git clone https://github.com/MCG-NJU/MOC-Detector.git ${MOC_ROOT}

  4. Install the requirements

    pip install -r pip-list.txt

  5. Compile deformable convolutional in DLA backbone follow CenterNet.

    cd ${MOC_ROOT}/src/network/DCNv2
bash make.sh