github地址：https://github.com/niki-amini-naieni/CountGD

本次示例在windows WSL2中执行

1、下载安装

打开WSL命令行，首先查看是否安装了gcc

~/CountGD$ gcc --version
gcc (Ubuntu 11.4.0-1ubuntu1~22.04.2) 11.4.0
Copyright (C) 2021 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

若未安装执行：

sudo apt update
sudo apt install build-essential

再依次执行（anaconda3）

git clone https://github.com/niki-amini-naieni/CountGD.git
conda create -n countgd python=3.9.19
conda activate countgd
cd CountGD
pip install -r requirements.txt
export CC=/usr/bin/gcc-11 # this ensures that gcc 11 is being used for compilation
cd models/GroundingDINO/ops
python setup.py build install
python test.py # should result in 6 lines of * True
pip install git+https://github.com/facebookresearch/segment-anything.git
cd ../../../

若 pip install git+https://github.com/facebookresearch/segment-anything.git下载很慢，可首先下载再安装

~/CountGD$ git clone https://github.com/facebookresearch/segment-anything.git
~/CountGD$ cd segment-anything
~/CountGD/segment-anything$ pip install .

下载bert-base-uncased

~/CountGD/segment-anything$ cd ..
~/CountGD$ mkdir checkpoints
~/CountGD$ python download_bert.py

下载groundingdino_swinb_cogcoor.pth

~/CountGD$ wget -P checkpoints https://github.com/IDEA-Research/GroundingDINO/releases/download/v0.1.0-alpha2/groundingdino_swinb_cogcoor.pth

下载sam_vit_h_4b8939.pth

~/CountGD$ wget -P checkpoints https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth

2、预测推理

下载预训练权重checkpoint_fsc147_best.pth：https://drive.google.com/file/d/1RbRcNLsOfeEbx6u39pBehqsgQiexHHrI/view

放到checkpoints文件夹

单张图片预测（single_image_inference.py）

~/CountGD$python single_image_inference.py --pretrain_model_path checkpoints/checkpoint_fsc147_best.pth --image_path test.png --output_image_name "test2.jpg" --text 'people'

--pretrain_model_path checkpoints/checkpoint_fsc147_best.pth (预训练权重)

--image_path test.png(待推理的图片)

--output_image_name "test2.jpg" (保存结果)

--text 'people' (提示词)

测试效果

--text 'people'

--text 'car'

--text '真理'    (missile

也可训练自己的数据

参考：https://github.com/niki-amini-naieni/CountGD/blob/main/training.md