OmniParser
安装Conda
conda
conda create -n "omni" python==3.12
conda activate omni安装torch
conda
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121克隆仓库，下载模型，模型放至 ./weights
bash
git clone https://github.com/microsoft/OmniParser.git
cd OmniParser
pip install -r requirements.txtpython
import os
from huggingface_hub import snapshot_download
def download_omni_v2():
print("🚀 正在检查并下载 OmniParser V2 权重...")
# 只下载你需要的 V2.0 权重
snapshot_download(
repo_id="microsoft/OmniParser-v2.0",
local_dir="weights",
allow_patterns=["icon_detect/*", "icon_caption/*"]
)
# 按照项目要求重命名目录（这是原项目的一个小坑）
old_path = os.path.join("weights", "icon_caption")
new_path = os.path.join("weights", "icon_caption_florence")
if os.path.exists(old_path) and not os.path.exists(new_path):
os.rename(old_path, new_path)
print("✅ 权重目录重命名完成")
print("✨ 所有模型已就绪！路径：./weights")
if __name__ == "__main__":
download_omni_v2()只启动OmniParser项目的server api
bash
cd omnitool/omniparserserver
python -m omniparserserver跑通后会发现omniparser无法解析中文
找到./util/utils.py修改如下代码：
python
from matplotlib import pyplot as plt
import easyocr
from paddleocr import PaddleOCR
reader = easyocr.Reader(['ch_sim', 'en']) #中英混合
paddle_ocr = PaddleOCR(
lang='ch', # 中文OCR模型
use_angle_cls=False,
use_gpu=False, # using cuda will conflict with pytorch in the same process
show_log=False,
max_batch_size=1024,
use_dilation=True, # improves accuracy
det_db_score_mode='slow', # improves accuracy
rec_batch_num=1024)- 我们搜索的是
坝= 0x574d - OmniParser 返回的是
坝= 0x575d
Contributors
root