Detectron2 Packages

FAIR Detectron2 https://github.com/facebookresearch/detectron2

Model Zoo Backbone/Pretrained are saved in ~/.torch/fvcore_cache/detectron2

Trained Models

https://drive.google.com/drive/folders/1IkcKH8hAhpjryCaGts3VwGPuOnlKRJMe?usp=sharing

Colab Notebooks

COCO 2017 Training

COCO 2017 Colab Download/Setup

COCO Download & Setup

Setup and Install

Detectron2 GPU

# Clone and go into directory
git clone --recurse-submodules https://github.com/sawyermade/detectron2_pkgs.git
cd detectron2_pkgs

# Create conda environment and activate it
conda create -n detectron2 python=3.6 -y
conda activate detectron2

# Install packages
conda install pytorch torchvision cudatoolkit=10.1 opencv -c pytorch -y
pip install -U cython requests jsonpickle pyrealsense2 flask imageio shapely flask-ngrok 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'
#pip install -U 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'
pip install -U 'git+https://github.com/cocodataset/panopticapi.git'
pip install -U 'git+https://github.com/lvis-dataset/lvis-api.git'
pip install -U 'git+https://github.com/mcordts/cityscapesScripts.git'
rm -rf detectron2/build detectron2/**/*.so
pip install -e detectron2

Detectron2 CPU

conda create -n detectron2cpu python=3.6 -y
conda activate detectron2cpu
pip install detectron2 -f \
	https://dl.fbaipublicfiles.com/detectron2/wheels/cpu/index.html
conda install pytorch torchvision cpuonly -c pytorch
conda install cython -y
conda install -c menpo opencv -y
pip install 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'
pip install requests jsonpickle pyrealsense2 flask

Packages

Demos

Go to demo directory

HTTP Server

Go to http directory

ROS Package

Go to ros directory

Training

Go to train directory