1: Install Annaconda
2: Install pycharm
3: Create a Python3.8 environment
conda create -n py38 python=3.8
4: Install GCC
Refer to the boss here: (61 messages) nhyltt's blog_CSDN blog-C language level two exam questions, a C language programming question every day, blogger in the field of notes
1. Install MinGW
Download MinGW Open the hosting website sourceforge and download
After the download is complete, double-click to open and click Install, all the way to Next installation
2. Install the GCC compiler
Open the compiler and click Basic Setup, find mingw-gcc-g++ -bin in Packpag, left-click the small box and click Mark for installation
Then click Installation in the menu bar, click Apply Changes
Click Apply in the pop-up dialog box, and then wait for the installation to complete
3. Configure environment variables
First find the path of the MinGW installation bin file, and then copy
Go back and click the small circle on the lower left to search, search for environment variables, and click to edit system environment variables
Then click on Environment Variables
Then click Path first, then click Edit
First click New, then add the copied path in the new box, and then click OK, and click OK in the previous dialog box
After clicking, the installation is normally completed, test it immediately
Press the WIN+R key, then enter cmd, then enter gcc -v, if the following information appears, the installation is successful
5: Install pytorch, pay attention to the corresponding cuda version
Go to pytorch.org
Install after selecting the correct torch.
5: Install the library
pip install -U opencv-python
conda install git
pip install fvcore
6: install detectron2
git clone [email protected]:facebookresearch/detectron2.git
cd detectron2
pip install -e .
pip install git+https://github.com/cocodataset/panopticapi.git
pip install git+https://github.com/mcordts/cityscapesScripts.git
pip install pycocotools-windows
7: Install dependent library files
pip install -r requirements.txt
8: run
cd mask2former/modeling/pixel_decoder/ops
sh make.sh
Note: Visual studio and cuda need to be installed