TypeError：图像数据无法转换为浮点数图像数据无法转换为浮点数

Question

当我尝试使用开放的cv 和flask 模块处理python 以从图像中删除背景颜色时。在运行代码时，我收到此错误：

File "new.py", line 82, in <module>
    plt.imshow(None)
  File "C:\Users\USER\Anaconda3\lib\site-packages\matplotlib\pyplot.py", line 2699, in imshow
    None else {}), **kwargs)
  File "C:\Users\USER\Anaconda3\lib\site-packages\matplotlib\__init__.py", line 1810, in inner
    return func(ax, *args, **kwargs)
  File "C:\Users\USER\Anaconda3\lib\site-packages\matplotlib\axes\_axes.py", line 5494, in imshow
    im.set_data(X)
  File "C:\Users\USER\Anaconda3\lib\site-packages\matplotlib\image.py", line 634, in set_data
    raise TypeError("Image data cannot be converted to float")
TypeError: Image data cannot be converted to float

我不知道如何解决这个问题。还有我正在工作的代码：

import io, traceback

from flask import Flask, request, g
from flask import send_file
from flask_mako import MakoTemplates, render_template
from plim import preprocessor
import matplotlib.pyplot as plt
from PIL import Image, ExifTags
from scipy.misc import imresize
import numpy as np
import cv2
from keras.models import load_model
import tensorflow as tf

app = Flask(__name__, instance_relative_config=True)
# For Plim templates
mako = MakoTemplates(app)
app.config['MAKO_PREPROCESSOR'] = preprocessor
app.config.from_object('config')
image= cv2.imread("1.jpg")
graph = tf.get_default_graph()

def ml_predict(image):
    with graph.as_default():
       # Add a dimension for the batch
       prediction = img.predict(image[None, :, :, :])
       prediction = prediction.reshape((224,224, -1))
       return prediction

def rotate_by_exif(image):
    try:
       for orientation in ExifTags.TAGS.keys() :
           if ExifTags.TAGS[orientation]=='Orientation' : break
           exif=dict(image._getexif().items())
           if not orientation in exif:
               return image

           if exif[orientation] == 3 :
               image=image.rotate(180, expand=True)
           elif exif[orientation] == 6 :
               image=image.rotate(270, expand=True)
           elif exif[orientation] == 8 :
               image=image.rotate(90, expand=True)
           return image
    except:
        traceback.print_exc()
        return image

THRESHOLD = 0.5

def predict():
    # Load image
    #image = request.files['file']
    image = Image.open(image)
    image = rotate_by_exif(image)
    resized_image = imresize(image, (224, 224)) / 255.0
    # Model input shape = (224,224,3)
    # [0:3] - Take only the first 3 RGB channels and drop ALPHA 4th channel in case this is a PNG
    prediction = ml_predict(resized_image[:, :, 0:3])
    print('PREDICTION COUNT', (prediction[:, :, 1]>0.5).sum())

    # Resize back to original image size
    # [:, :, 1] = Take predicted class 1 - currently in our model = Person class. Class 0 = Background
    prediction = imresize(prediction[:, :, 1], (image.height, image.width))
    prediction[prediction>THRESHOLD*255] = 255
    prediction[prediction<THRESHOLD*255] = 0

    # Append transparency 4th channel to the 3 RGB image channels.
    transparent_image = np.append(np.array(image)[:, :, 0:3], prediction[: , :, None], axis=-1)
    transparent_image = Image.fromarray(transparent_image)



plt.imshow(None)
plt.show()

非常感谢任何有关此的帮助。并预先感谢您帮助解决此问题。

Answer 1

您需要在predict 函数中调用plt.imshow(transparent_image)。

它抱怨 transparent_image 未定义，因为您试图在其范围之外使用 predict 函数的局部变量。