docker 环境部署

.gitignore

@@ -0,0 +1,5 @@
+__pycache__
+.vscode
+.DS_Store
+.coverage
+output

Dockerfile

@@ -0,0 +1,30 @@
+# https://huggingface.co/spaces/gaunernst/layoutlm-docvqa-paddleocr/blob/main/Dockerfile
+FROM python:3.10
+
+ENV CUDA_VISIBLE_DEVICES=-1
+ARG PIP_NO_CACHE_DIR=1
+
+# libssl1.1 for PaddlePaddle
+RUN wget http://security.ubuntu.com/ubuntu/pool/main/o/openssl/libssl1.1_1.1.1f-1ubuntu2.20_amd64.deb \
+    && dpkg -i libssl1.1_1.1.1f-1ubuntu2.20_amd64.deb \
+    && rm libssl1.1_1.1.1f-1ubuntu2.20_amd64.deb
+
+RUN pip install torch==2.1.1 -i https://download.pytorch.org/whl/cpu
+RUN pip install paddlepaddle==2.5.1 -i https://mirror.baidu.com/pypi/simple
+RUN pip install transformers gradio Pillow fastapi
+RUN pip install paddleocr==2.7.0.3 \
+    && pip uninstall -y opencv-python opencv-contrib-python \
+    && pip install opencv-python-headless
+
+RUN useradd -m -u 1000 user
+USER user
+ENV HOME=/home/user
+WORKDIR $HOME/app
+RUN mkdir output
+COPY --chown=user app.py $HOME/app
+COPY --chown=user table.jpg $HOME/app 
+COPY --chown=user word_1.jpg $HOME/app
+RUN paddleocr --image_dir ./word_1.jpg --use_angle_cls true --use_gpu false  --lang ch
+RUN paddleocr --image_dir ./table.jpg --type=structure --table=true  --lang ch
+
+CMD ["python", "-u", "app.py"]

app.py

@@ -0,0 +1,89 @@
+import uvicorn
+from fastapi.staticfiles import StaticFiles
+import hashlib
+from enum import Enum
+from fastapi import FastAPI, UploadFile, File
+from paddleocr import PaddleOCR, PPStructure, save_structure_res
+from PIL import Image
+import io
+import numpy as np
+
+app = FastAPI()
+use_gpu = False
+output_dir = 'output'
+
+class LangEnum(str, Enum):
+    ch = "ch"
+    en = "en"
+
+# cache with ocr
+ocr_cache = {}
+
+# get ocr ins
+def get_ocr(lang, use_gpu=False):
+    if not ocr_cache.get(lang):
+        ocr_cache[lang] = PaddleOCR(use_angle_cls=True, lang=lang, use_gpu=use_gpu)
+    
+    return ocr_cache.get(lang)
+
+
+@app.post("/ocr")
+async def create_upload_file(
+    file: UploadFile = File(...),
+    lang: LangEnum = LangEnum.ch,
+    # use_gpu: bool = False
+):
+    contents = await file.read()
+    image = Image.open(io.BytesIO(contents))
+    ocr = get_ocr(lang=lang, use_gpu=use_gpu)
+    img2np = np.array(image)
+    result = ocr.ocr(img2np, cls=True)[0]
+
+    boxes = [line[0] for line in result]
+    txts = [line[1][0] for line in result]
+    scores = [line[1][1] for line in result]
+
+    # 识别结果
+    final_result = [dict(boxes=box, txt=txt, score=score) for box, txt, score in zip(boxes, txts, scores)]
+    return final_result
+
+
+@app.post("/ocr_table")
+async def create_upload_file(
+    file: UploadFile = File(...),
+    lang: LangEnum = LangEnum.ch,
+    # use_gpu: bool = False
+):
+    table_engine = PPStructure(show_log=True, table=True, lang=lang)
+
+    contents = await file.read()
+    # 计算文件内容的哈希值
+    file_hash = hashlib.sha256(contents).hexdigest()
+
+    image = Image.open(io.BytesIO(contents))
+    img2np = np.array(image)
+    result = table_engine(img2np)
+
+    save_structure_res(result, output_dir, f'{file_hash}')
+
+    htmls = []
+    types = []
+    bboxes = []
+
+    for item in result:
+        item_res = item.get('res', {})
+        htmls.append(item_res.get('html', ''))
+        types.append(item.get('type', ''))
+        bboxes.append(item.get('bbox', ''))
+
+    return {
+        'htmls': htmls,
+        'hash': file_hash,
+        'bboxes': bboxes,
+        'types': types,
+    }
+
+
+if __name__ == '__main__':
+    app.mount("/output", StaticFiles(directory="output", follow_symlink=True, html=True), name="output")
+    uvicorn.run(app=app, port=7860)