Setup Detectron2 on Cloud Linux

After setting up CUDA on Machines, we will install detectron 2 now.

1. Setup the environment

I would recommend using conda to manage environments

If you don’t have Anaconda/Miniconda instaled on your system, Install it using the following steps:-

a) Download Miniconda

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

b) Change permission and make the file executable

chmod 777 Miniconda3-latest-Linux-x86_64.sh

c) Run the Installer and follow instructions

./Miniconda3-latest-Linux-x86_64.sh

d) Reload ~/.bashrc

source .bashrc

Now that we have conda installed, we can create virtual environments.

Create a virtual environment

conda create -n detectron2_env

Activate the Virtual Environment.

conda activate detectron2_env

Step 2: Install Requirements

Install pip if not installed

conda install pip

Make a requirement.txt file containing requirements for detectron2, and copy the following content in requirement.txt

opencv-python
torch
torchvision 
git+https://github.com/facebookresearch/fvcore

then install the requirements by:

pip install -r detectron_requirement.txt

Step 3: Install detectron2

Install detectron2 by running these commands

git clone https://github.com/facebookresearch/detectron2 detectron2_repo
pip install --user -e detectron2_repo

Step 4: Check the installation

To check if detectron2 is successfully installed run these commands

cd detectron2_repo
wget http://images.cocodataset.org/val2017/000000439715.jpg -O input.jpg
python3 demo/demo.py --config-file configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml --input input.jpg --output outputs --opts MODEL.WEIGHTS detectron2://COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x/137849600/model_final_f10217.pkl

If all commands runs without errors, Hooray, Detectron2 is setup on your system, you can now use it to train the model and make inferences.