recognize-anything 识别万物

docker run --gpus all -itd --name=test -v /app:/app nvcr.io/nvidia/cuda:11.8.0-devel-ubuntu22.04

###########安装过程########################

更新包索引

apt update

安装 Python 3

apt install python3 -y

安装 pip

apt install python3-pip -y

安装 Git

apt install git -y

安装 ffmpeg 和一些图形相关的库

apt-get install ffmpeg libsm6 libxext6 -y

安装 OpenGL 库

apt-get install libgl1 -y

将 Python 3 的可执行文件复制到一个更通用的命令名

cp /usr/bin/python3 /usr/bin/python

安装 wget

apt-get install -y wget

克隆 "recognize-anything" 项目

git clone https://github.com/xinyu1205/recognize-anything.git

进入项目文件夹

cd recognize-anything/

安装一些 Python 包

pip install timm transformers fairscale pycocoevalcap

安装项目所需的所有依赖项

pip install -r requirements.txt

创建并进入一个名为 `pretrained` 的文件夹

mkdir pretrained

cd pretrained

使用 wget 下载预训练模型

wget https://huggingface.co/xinyu1205/recognize-anything-plus-model/resolve/main/ram_plus_swin_large_14m.pth

wget https://huggingface.co/xinyu1205/recognize-anything-plus-model/resolve/main/ram_plus_tag_embedding_class_4585_des_51.pth

wget https://huggingface.co/spaces/xinyu1205/Recognize_Anything-Tag2Text/resolve/main/ram_swin_large_14m.pth

wget https://huggingface.co/spaces/xinyu1205/Recognize_Anything-Tag2Text/resolve/main/tag2text_swin_14m.pth

#######################

启动脚本如下

python inference_ram_plus.py --image images/demo/demo1.jpg \

> --pretrained pretrained/ram_plus_swin_large_14m.pth