Paddle-Lite 树莓派4B+ python 部署 yolov3模型出错

uelo1irk  于 2022-04-21  发布在  Java
关注(0)|答案(1)|浏览(474)

树莓派4B+ python 3.7.3

利用yolov3 【608,608】 生成 model.nb

执行如下程序

from paddlelite.lite import *
from PIL import Image
import numpy as np

设置config信息

config = MobileConfig()
config.set_model_from_file("model.np")

创建predictor

predictor = create_paddle_predictor(config)

从图片读入数据

image = Image.open('test.jpg')
resized_image = image.resize((608, 608), Image.BILINEAR)
image_data = np.array(resized_image).flatten().tolist()

设置输入数据

input_tensor = predictor.get_input(0)
input_tensor.resize([1, 3, 608 608])
input_tensor.set_float_data(image_data)

执行预测

predictor.run()

得到输出数据

output_tensor = predictor.get_output(0)

print(output_tensor.shape())
print(output_tensor.float_data()[:10])

错误代码

Backend terminated or disconnected.Fatal Python error: Segmentation fault

Current thread 0xb6fafad0 (most recent call first):

File "/home/pi/Desktop/run_jpg.py", line 27 in
File "/usr/lib/python3/dist-packages/thonny/backend.py", line 1272 in _execute_prepared_user_code
File "/usr/lib/python3/dist-packages/thonny/backend.py", line 1200 in wrapper
File "/usr/lib/python3/dist-packages/thonny/backend.py", line 1213 in wrapper
File "/usr/lib/python3/dist-packages/thonny/backend.py", line 1259 in execute_source

问题1:yolov3 如何 设置输入数据
问题2:从图片如何读入数据

wkyowqbh

wkyowqbh1#

yolov3需要两个输入tensor,另一个input_tensor1 = predictor.get_input(1),你需要设置图像尺寸。

相关问题