未识别图像错误：无法识别图像文件<_io.BytesIO object at 0x000002154C6AE400>。

Question

我正在尝试使用预训练的CNN模型(densenet121)的保存权重来测试其准确性，但遇到了以下错误。测试的代码如下：

image_size=512
batch_size=4
model_name='weights/densenet121_CAB_messidor.h5' #pretrained weight of the model
test_dir='./data/Messidor_Classes/'   #dataset directory
  
model=load_model(model_name) #loading the model

test_num=0
result=np.zeros((4,4),dtype=int)
recall=np.zeros((1,4),dtype=float)
for i in range(4): #one loop for each class
        datadirs=test_dir+str(i)+'/' #image directory
        filenames=os.listdir(datadirs)
        num=len(filenames)
        test_num=test_num+num
        valid = ImageDataGenerator() #data augmentation
        valid_data=valid.flow_from_directory(directory=test_dir,target_size=(image_size,image_size),
                                             batch_size=batch_size,class_mode=None,classes=str(i))
        predict=model.predict(valid_data,batch_size=batch_size,verbose=1,workers=1) #the line where the error is encountered
        predict=np.argmax(predict,axis=-1)
        for j in range(4):
            result[i,j]=np.sum(predict==j)

right=result[0,0]+result[1,1]+result[2,2]+result[3,3]
print('Acc:',right/test_num) #accuracy

w_kappa=weight_kappa(result,test_num)
print('w_kappa:',w_kappa) #kappa score

该数据集的树状结构如下，图像格式为.tif：

 Messidor--|0->img_1.tif #images for class 0
           |   ...
           |   ...
           | ->img_2.tif 
           |1->img_1.tif #images for class 1
           |   ...
           |   ...
           | ->img_2.tif
           |2->img_1.tif #images for class 2
           |   ...
           |   ...
           | ->img_2.tif
           |3->img_1.tif #images for class 3
           |   ...
           |   ...
           | ->img_2.tif

运行这个测试代码时我遇到的错误是：

Found 428 images belonging to 1 classes.
Traceback (most recent call last):

  File "C:\Users\aokra\source\repos\CABnet\test.py", line 53, in <module>
    predict=model.predict(valid_data,batch_size=batch_size,verbose=1,workers=1)

  File "C:\Users\aokra\anaconda3\lib\site-packages\tensorflow\python\keras\engine\training.py", line 130, in _method_wrapper
    return method(self, *args, **kwargs)

  File "C:\Users\aokra\anaconda3\lib\site-packages\keras_preprocessing\image\iterator.py", line 227, in _get_batches_of_transformed_samples
    img = load_img(filepaths[j],

  File "C:\Users\aokra\anaconda3\lib\site-packages\keras_preprocessing\image\utils.py", line 114, in load_img
    img = pil_image.open(io.BytesIO(f.read()))

  File "C:\Users\aokra\anaconda3\lib\site-packages\PIL\Image.py", line 2967, in open
    raise UnidentifiedImageError(

UnidentifiedImageError: cannot identify image file <_io.BytesIO object at 0x000001C299030EF0>


   

- arvind okram

Answer 1

此错误是由于损坏的文件或文件扩展名不正确，生成器无法读取。
使用以下函数检查输入数据：
代码：

import os
from PIL import Image
folder_path = r'data\Messidor_Classes'
extensions = []
for fldr in os.listdir(folder_path):
    sub_folder_path = os.path.join(folder_path, fldr)
    for filee in os.listdir(sub_folder_path):
        file_path = os.path.join(sub_folder_path, filee)
        print('** Path: {}  **'.format(file_path), end="\r", flush=True)
        im = Image.open(file_path)
        rgb_im = im.convert('RGB')
        if filee.split('.')[1] not in extensions:
            extensions.append(filee.split('.')[1])

- Aniket Bote

---

非常感谢您的快速回复和帮助，如果您能再帮我一次就太好了。我已经运行了您给出的代码。当我运行它时，输出结果如下："UnidentifiedImageError: cannot identify image file 'Messidor_Classes\0\20051020_43808_0100_PP.tif'"。这是不是意味着图像文件已经损坏了？ - arvind okram

第10行显示错误：“im = Image.open(file_path)”。 - arvind okram

是的。如果您收到错误消息，则表示TensorFlow无法读取打印路径中存在的图像。 - Aniket Bote

问题已经解决。谢谢。 - arvind okram

非常感谢您提供的代码！它帮我节省了很多工作。 - fattah.safa