seal-recognition 服务器调用异常

[linuxlin365]

描述问题

复现

1. 运行：paddlex --serve --pipeline seal_recognition
2. 根据py调用示例，调用

import base64
import requests

API_URL = "http://10.26.18.66:8080/seal-recognition" # 服务URL
image_path = "test3.png"


# 对本地图像进行Base64编码
with open(image_path, "rb") as file:
    image_bytes = file.read()
    image_data = base64.b64encode(image_bytes).decode("ascii")

payload = {"image": image_data}  # Base64编码的文件内容或者图像URL

# 调用API
response = requests.post(API_URL, json=payload)

# 处理接口返回数据
assert response.status_code == 200
result = response.json()["result"]
with open(layout_image_path, "wb") as file:
    file.write(base64.b64decode(result["layoutImage"]))
print(f"Output image saved at {layout_image_path}")
print("\nDetected seal impressions:")
print(result["sealImpressions"])

3. 服务器端报错：

{'input_path': PosixPath('/root/.paddlex/temp/tmpz6d183hq.png'), 'layout_result': {'input_path': PosixPath('/root/.paddlex/temp/tmpz6d183hq.png'), 'boxes': [{'cls_id': 2, 'label': 'seal', 'score': 0.9772647023200989, 'coordinate': [118.36611, 86.67954, 914.26227, 885.9579]}]}, 'ocr_result': {'dt_polys': [array([[578, 158],
       [582, 159],
       [733, 207],
       [739, 210],
       [746, 217],
       [827, 334],
       [830, 340],
       [855, 411],
       [857, 419],
       [860, 512],
       [859, 520],
       [821, 656],
       [818, 663],
       [812, 669],
       [806, 673],
       [800, 675],
       [795, 676],
       [772, 676],
       [765, 675],
       [760, 673],
       [689, 635],
       [683, 631],
       [679, 625],
       [676, 618],
       [675, 610],
       [676, 605],
       [703, 488],
       [653, 373],
       [568, 327],
       [514, 313],
       [458, 337],
       [405, 378],
       [359, 481],
       [357, 503],
       [409, 630],
       [411, 637],
       [410, 645],
       [408, 652],
       [404, 658],
       [399, 663],
       [392, 667],
       [387, 668],
       [263, 683],
       [256, 683],
       [249, 680],
       [242, 677],
       [236, 669],
       [184, 565],
       [182, 558],
       [175, 525],
       [175, 514],
       [196, 404],
       [197, 400],
       [235, 297],
       [240, 288],
       [316, 206],
       [322, 201],
       [328, 198],
       [504, 149],
       [511, 148],
       [515, 148]])], 'dt_scores': [0.9914360869106089], 'rec_text': ['上海科技有限公司'], 'rec_score': [0.9992580413818359], 'input_path': PosixPath('/root/.paddlex/temp/tmpz6d183hq.png')}, 'src_file_name': array([[[255, 255, 255],
        [255, 255, 255],
        [255, 255, 255],
        ...,
        [255, 255, 255],
        [255, 255, 255],
        [255, 255, 255]],

       [[255, 255, 255],
        [255, 255, 255],
        [255, 255, 255],
        ...,
        [255, 255, 255],
        [255, 255, 255],
        [255, 255, 255]],

       [[255, 255, 255],
        [255, 255, 255],
        [255, 255, 255],
        ...,
        [255, 255, 255],
        [255, 255, 255],
        [255, 255, 255]],

       ...,

       [[255, 255, 255],
        [255, 255, 255],
        [255, 255, 255],
        ...,
        [255, 255, 255],
        [255, 255, 255],
        [255, 255, 255]],

       [[255, 255, 255],
        [255, 255, 255],
        [255, 255, 255],
        ...,
        [255, 255, 255],
        [255, 255, 255],
        [255, 255, 255]],

       [[255, 255, 255],
        [255, 255, 255],
        [255, 255, 255],
        ...,
        [255, 255, 255],
        [255, 255, 255],
        [255, 255, 255]]], dtype=uint8), 'page_id': 0}
string indices must be integers
Traceback (most recent call last):
  File "/root/PaddleX/paddlex/inference/pipelines/serving/_pipeline_apps/seal_recognition.py", line 88, in _infer
    item["dt_polys"], item["rec_text"], item["rec_score"]
TypeError: string indices must be integers
INFO:     10.26.16.57:50861 - "POST /seal-recognition HTTP/1.1" 500 Internal Server Error

环境

1. 请提供您使用的PaddlePaddle、PaddleX版本号、Python版本号
   PaddlePaddle 3.0.0-beta1
   Python 3.10.15

2. 请提供您使用的操作系统信息，如Linux/Windows/MacOS
   linux

3. 请问您使用的CUDA/cuDNN的版本号是？
   无

[Bobholamovic]

你好，这是3.0.0-beta1版本的一个bug，请使用3.0.0-beta2版本的paddlex试试～