如何更改 VGG16 中的批量大小？

Question

如何更改 VGG16 中的批量大小？ 我试图通过这样做来解决超过 memory 约束 10% 的问题。

错误：

2021-12-03 16:17:07.263665: W tensorflow/core/framework/cpu_allocator_impl.cc:82] Allocation of 4888553472 exceeds 10% of free system memory.

这是我的代码：

def labelObjectFromImage(image_path, directory_filename):
    img = cv2.imread(image_path+directory_filename)
    height = img.shape[0]
    width = img.shape[1]
    channels = img.shape[2]
    img = load_img(image_path+directory_filename, target_size=(height, width))
    model = VGG16(weights="imagenet", include_top = False, input_shape = (height, width, channels))
    img = img_to_array(img)
    img = img.reshape((1, img.shape[0], img.shape[1], img.shape[2]))
    img = preprocess_input(img)
    yhat = model.predict(img)
    label = decode_predictions(yhat)
    label = label[0][0]
    print(label)

我尝试将 model.predict 更改为：

    yhat = model.predict(img, batch_size=1)

但它似乎对尝试解决问题没有任何影响

我尝试使用：

from tensorflow.keras import backend as K

K.clear_session()

但这并没有帮助

我跑了

pip3 uninstall tensorflow-gpu

然后通过安装正常的 tensorflow

pip3 install tensorflow

但这并没有帮助

仅供参考，到目前为止，我在所有这些尝试中都遇到了同样的错误。

正如建议的那样，我尝试了：

    img_resized = tf.image.resize(img, (height, width))

但我现在收到以下错误：

Traceback (most recent call last):
  File "organizeSpreadsheet.py", line 105, in <module>
    main()
  File "organizeSpreadsheet.py", line 86, in main
    objects_from_image = labelObjectFromImage(path_to_images, directory_filename)
  File "organizeSpreadsheet.py", line 53, in labelObjectFromImage
    img = img_resized.reshape((1, height, width, channels))
  File "/home/jr/.local/lib/python3.8/site-packages/tensorflow/python/framework/ops.py", line 437, in __getattr__
    raise AttributeError("""
AttributeError: 
        'EagerTensor' object has no attribute 'reshape'.
        If you are looking for numpy-related methods, please run the following:
        from tensorflow.python.ops.numpy_ops import np_config
        np_config.enable_numpy_behavior()

我知道我没有完全按照指示做，但它抛出了一个错误，所以我遵循了这个建议

我通过进行以下更改更正了错误：

def labelObjectFromImage(image_path, directory_filename):
    scale = 60

    img = cv2.imread(image_path+directory_filename)
    height = int(img.shape[0] * scale / 100)
    width = int(img.shape[1] * scale / 100)
    channels = img.shape[2]
    #img_resized = cv2.resize(img, dim, interpolation=cv2.INTER_AREA)
    #img_resized = tf.image.resize(img, (height, width))
    tf.image.resize(img, (height, width))
    #img = load_img(image_path+directory_filename, target_size=(height, width))
    model = VGG16(weights="imagenet", include_top = False, input_shape = (height, width, channels))
    #img = img_to_array(img)
    #img = img.reshape((1, img.shape[0], img.shape[1], img.shape[2]))
    img = img.reshape((1, height, width, channels))
    img = preprocess_input(img)
    yhat = model.predict(img, batch_size=1)
    label = decode_predictions(yhat)
    label = label[0][0]
    print(label)

但现在我得到了错误：

ValueError: cannot reshape array of size 63483840 into shape (1,3384,2251,3)

我认为这可以通过尝试多个尺度来解决，对吗？

Answer 1

您已经在使用batch_size = 1。

1. 通过在导入 tensorflow 时检查日志来检查您是否正在使用 gpu。
2. 在使用tf.image.resize(image, [small_height,small_width,N_channels])进行预测之前尝试调整图像大小