Skip to content

Latest commit

 

History

History
60 lines (42 loc) · 1.52 KB

INSTALL.md

File metadata and controls

60 lines (42 loc) · 1.52 KB
Raw

Installation

The code was tested on Ubuntu 18.04 with Cuda 10.0, Anaconda Python 3.7 and PyTorch v1.4.0. NVIDIA GPUs are needed for both training and testing. After installing Anaconda:

  1. [Optional but recommended] create a new conda environment.

    conda create --name HoughNet python=3.7

    And activate the environment.

    conda activate HoughNet

  2. Clone the repo:

    HoughNet_ROOT=/path/to/clone/HoughNet
git clone https://github.com/nerminsamet/HoughNet $HoughNet_ROOT

  3. Install PyTorch 1.4.0:

    conda install pytorch torchvision cudatoolkit=10.0 -c pytorch

  4. Install the requirements:

    pip install -r requirements.txt

  5. Install Detectron for instance segmentation task:

    python -m pip install detectron2==0.2.1 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cu100/torch1.4/index.html

  6. Compile DCNv2 (Deformable Convolutional Networks):

    cd $HoughNet_ROOT/src/lib/models/networks/DCNv2
./make.sh

  7. [Optional, only required if you are using multi-scale testing]. Compile NMS if you want to use multi-scale testing.

    cd $HoughNet_ROOT/src/lib/external
make

  8. Download pretrained models and place them under $HoughNet_ROOT/models/. You could find more information about models in model zoo.