我已经使用mnist数据集使用keras构建了一个神经网络,现在我正尝试在实际手写数字的照片上使用它。当然,我并不期望结果会是完美的,但是我目前得到的结果仍有很大的改进空间。
对于初学者来说,我会用我最清晰的笔迹写的一些个人数字照片进行测试。它们是正方形的,并且具有与mnist数据集中的图像相同的尺寸和颜色。它们被保存在名为 individual_test 的文件夹中,例如: 7(2)_digit.jpg 。
网络经常非常确定错误的结果,我将举一个例子:
我为这张照片得到的结果如下:
result: 3 . probabilities: [1.9963557196245318e-10, 7.241294497362105e-07, 0.02658148668706417, 0.9726449251174927, 2.5416460047722467e-08, 2.6078915027483163e-08, 0.00019745019380934536, 4.8302300825753264e-08, 0.0005754049634560943, 2.8358477788259506e-09]
因此,网络有97%的人确定这是3,而这并不是唯一的情况。在38张照片中,只有16张被正确识别。令我感到震惊的是,尽管网络离正确的结果再远了,但它对结果的把握如此确定。
我该怎么做才能提高性能?我可以更好地准备图像吗?还是应该将自己的图像添加到训练数据中?如果是这样,我该怎么做?
编辑
这是在上面应用 prepare_image 后上面显示的图片的样子:
相比:这是mnist数据集提供的图片之一:
它们看起来和我非常相似。我该如何改善呢?
这是我的代码:
# import keras and the MNIST dataset
from tensorflow.keras.datasets import mnist
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from keras.utils import np_utils
# numpy is necessary since keras uses numpy arrays
import numpy as np
# imports for pictures
import PIL
# imports for tests
import random
import os
class mnist_network():
def __init__(self):
""" load data, create and train model """
# load data
(X_train, y_train), (X_test, y_test) = mnist.load_data()
# flatten 28*28 images to a 784 vector for each image
num_pixels = X_train.shape[1] * X_train.shape[2]
X_train = X_train.reshape((X_train.shape[0], num_pixels)).astype('float32')
X_test = X_test.reshape((X_test.shape[0], num_pixels)).astype('float32')
# normalize inputs from 0-255 to 0-1
X_train = X_train / 255
X_test = X_test / 255
# one hot encode outputs
y_train = np_utils.to_categorical(y_train)
y_test = np_utils.to_categorical(y_test)
num_classes = y_test.shape[1]
# create model
self.model = Sequential()
self.model.add(Dense(num_pixels, input_dim=num_pixels, kernel_initializer='normal', activation='relu'))
self.model.add(Dense(num_classes, kernel_initializer='normal', activation='softmax'))
# Compile model
self.model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
# train the model
self.model.fit(X_train, y_train, validation_data=(X_test, y_test), epochs=10, batch_size=200, verbose=2)
self.train_img = X_train
self.train_res = y_train
self.test_img = X_test
self.test_res = y_test
def predict_result(self, img, num_pixels = None, show=False):
""" predicts the number in a picture (vector) """
assert type(img) == np.ndarray and img.shape == (784,)
"""if show:
# show the picture!!!! some problem here
plt.imshow(img, cmap='Greys')
plt.show()"""
num_pixels = img.shape[0]
# the actual number
res_number = np.argmax(self.model.predict(img.reshape(-1,num_pixels)), axis = 1)
# the probabilities
res_probabilities = self.model.predict(img.reshape(-1,num_pixels))
return (res_number[0], res_probabilities.tolist()[0]) # we only need the first element since they only have one
def prepare_image(self, img):
""" prepares the partial images used in partial_img_rec by transforming them
into numpy arrays that the network will be able to process """
# convert to greyscale
img = img.convert("L")
# rescale image to 28 *28 dimension
img = img.resize((28,28), PIL.Image.ANTIALIAS)
# inverse colors since the training images have a black background
img = PIL.ImageOps.invert(img)
# transform to vector
img = np.asarray(img, "float32")
img = img / 255.
img[img < 0.5] = 0.
# flatten image to 28*28 = 784 vector
num_pixels = img.shape[0] * img.shape[1]
img = img.reshape(num_pixels)
return img
def partial_img_rec(self, image, upper_left, lower_right, results=[], show = False):
""" partial is a part of an image """
left_x, left_y = upper_left
right_x, right_y = lower_right
print("current test part: ", upper_left, lower_right)
print("results: ", results)
# condition to stop recursion: we've reached the full width of the picture
width, height = image.size
if right_x > width:
return results
partial = image.crop((left_x, left_y, right_x, right_y))
if show:
partial.show()
partial = self.prepare_image(partial)
step = height // 10
# is there a number in this part of the image?
res, prop = self.predict_result(partial)
print("result: ", res, ". probabilities: ", prop)
# only count this result if the network is at least 50% sure
if prop[res] >= 0.5:
results.append(res)
# step is 80% of the partial image's size (which is equivalent to the original image's height)
step = int(height * 0.8)
print("found valid result")
else:
# if there is no number found we take smaller steps
step = height // 20
print("step: ", step)
# recursive call with modified positions ( move on step variables )
return self.partial_img_rec(image, (left_x + step, left_y), (right_x + step, right_y), results = results)
def individual_digits(self, img):
""" uses partial_img_rec to predict individual digits in square images """
assert type(img) == PIL.JpegImagePlugin.JpegImageFile or type(img) == PIL.PngImagePlugin.PngImageFile or type(img) == PIL.Image.Image
return self.partial_img_rec(img, (0,0), (img.size[0], img.size[1]), results=[])
def test_individual_digits(self):
""" test partial_img_rec with some individual digits (shape: square)
saved in the folder 'individual_test' following the pattern 'number_digit.jpg' """
cnt_right, cnt_wrong = 0,0
folder_content = os.listdir(".\individual_test")
for imageName in folder_content:
# image file must be a jpg or png
assert imageName[-4:] == ".jpg" or imageName[-4:] == ".png"
correct_res = int(imageName[0])
image = PIL.Image.open(".\\individual_test\\" + imageName).convert("L")
# only square images in this test
if image.size[0] != image.size[1]:
print(imageName, " has the wrong proportions: ", image.size,". It has to be a square.")
continue
predicted_res = self.individual_digits(image)
if predicted_res == []:
print("No prediction possible for ", imageName)
else:
predicted_res = predicted_res[0]
if predicted_res != correct_res:
print("error in partial_img-rec! Predicted ", predicted_res, ". The correct result would have been ", correct_res)
cnt_wrong += 1
else:
cnt_right += 1
print("correctly predicted ",imageName)
print(cnt_right, " out of ", cnt_right + cnt_wrong," digits were correctly recognised. The success rate is therefore ", (cnt_right / (cnt_right + cnt_wrong)) * 100," %.")
def multiple_digits(self, img):
""" takes as input an image without unnecessary whitespace surrounding the digits """
#assert type(img) == myImage
width, height = img.size
# start with the first square part of the image
res_list = self.partial_img_rec(img, (0,0),(height ,height), results = [])
res_str = ""
for elem in res_list:
res_str += str(elem)
return res_str
def test_multiple_digits(self):
""" tests the function 'multiple_digits' using some images saved in the folder 'multi_test'.
These images contain multiple handwritten digits without much whitespac surrounding them.
The correct solutions are saved in the files' names followed by the characte '_'. """
cnt_right, cnt_wrong = 0,0
folder_content = os.listdir(".\multi_test")
for imageName in folder_content:
# image file must be a jpg or png
assert imageName[-4:] == ".jpg" or imageName[-4:] == ".png"
image = PIL.Image.open(".\\multi_test\\" + imageName).convert("L")
correct_res = imageName.split("_")[0]
predicted_res = self.multiple_digits(image)
if correct_res == predicted_res:
cnt_right += 1
else:
cnt_wrong += 1
print("Error in multiple_digits! The network predicted ", predicted_res, " but the correct result would have been ", correct_res)
print("The network predicted correctly ", cnt_right, " out of ", cnt_right + cnt_wrong, " pictures. That's a success rate of ", cnt_right / (cnt_right + cnt_wrong) * 100, "%.")
network = mnist_network()
# 7(2).digit.jpg is the image shown above
network.individual_digits(PIL.Image.open(".\individual_test\\7(2)_digit.jpg"))
答案 0 :(得分:1)
您在MNIST数据集上的测试分数是多少? 我想到的一件事是您的图像缺少阈值,
阈值处理是将某个像素以下的像素值设置为零的技术,请参阅OpenCV阈值示例,您可能需要使用反阈值并再次检查结果。
做,通知是否有进展。
答案 1 :(得分:0)
您遇到的主要问题是,您正在测试的图像与MNIST图像不同,可能是由于已完成图像的准备工作,在您应用prepare_image之后,您能否显示来自正在测试的图像的图像?在上面。
答案 2 :(得分:0)
您可以通过以下三种方法在此特定任务中获得更好的性能:
我刚刚做了一个实验。我检查了每个代表一个数字的MNIST图像。我拍摄了您的图像,并进行了我之前向您建议的一些预处理,例如:
1。设置了一些阈值,但是只是向下消除了背景噪声,因为原始MNIST数据仅对空白背景具有一些最小阈值:
image[image < 0.1] = 0.
2。令人惊讶的是,图像内部数字的大小已被证明是至关重要的,因此我按比例缩放了28 x 28图像内部的数字。我们在数字周围还有更多填充。
3。。由于来自keras的MNIST数据也反转了,所以我反转了图像。
image = ImageOps.invert(image)
4。。最后,像我们在培训中一样,用来缩放数据:
image = image / 255.
预处理后,我使用MNIST数据集训练了模型,该数据集的参数为epochs=12, batch_size=200
,结果为:
结果: 1 ,概率: 0.6844741106033325
result: **1** . probabilities: [2.0584749904628552e-07, 0.9875971674919128, 5.821426839247579e-06, 4.979299319529673e-07, 0.012240586802363396, 1.1566483948399764e-07, 2.382085284580171e-08, 0.00013023221981711686, 9.620113416985987e-08, 2.5273093342548236e-05]
结果: 6 ,概率: 0.9221984148025513
result: 6 . probabilities: [9.130864782491699e-05, 1.8290626258021803e-07, 0.00020504613348748535, 2.1564576968557958e-07, 0.0002401985548203811, 0.04510130733251572, 0.9221984148025513, 1.9014490248991933e-07, 0.03216308355331421, 3.323434683011328e-08]
结果: 7 ,概率: 0.7105212807655334 注意:
result: 7 . probabilities: [1.0372193770535887e-08, 7.988557626958936e-06, 0.00031014863634482026, 0.0056108818389475346, 2.434678014751057e-09, 3.2280522077599016e-07, 1.4190952857262573e-09, 0.9940618872642517, 1.612859932720312e-06, 7.102244126144797e-06]
您的电话号码 9 有点棘手:
我发现带有MNIST数据集的模型获得了关于 9 的两个主要“特征”。上部和下部。与图像上一样,具有良好圆形的上部不是 9 ,而是针对MNIST数据集训练的模型的 3 。根据MNIST数据集, 9 的下部大部分是拉直曲线。因此,由于MNIST样本的缘故,基本上,理想形状的 9 始终是模型的 3 ,除非您再次用足够数量的形状的样本来训练模型> 9 。为了检查我的想法,我用 9 s做了一个子实验:
我的 9 的上部倾斜(根据MNIST,大多数情况下 9 都可以),但底部略微卷曲( 9 不能根据MNIST):
结果: 9 ,概率: 0.5365301370620728
我的 9 ,其上部倾斜(根据MNIST,大多数情况下 9 都可以),并且笔直的底部(按照 9 就可以了) MNIST):
结果: 9 ,概率: 0.923724353313446
您的 9 具有错误解释的形状属性:
结果: 3 ,概率: 0.8158268928527832
result: 3 . probabilities: [9.367801249027252e-05, 3.9978775021154433e-05, 0.0001467708352720365, 0.8158268928527832, 0.0005801069783046842, 0.04391581565141678, 6.44062723154093e-08, 7.099170943547506e-06, 0.09051419794559479, 0.048875387758016586]
最后只是证明图像缩放(填充)重要性的证据,我上面提到的很重要:
结果: 3 ,概率: 0.9845736622810364
结果: 9 ,概率: 0.923724353313446
因此我们可以看到,如果模型内部形状过大且填充尺寸较小,则模型会解释并解释为 3 。
我认为使用CNN可以获得更好的性能,但是采样和预处理方式对于在ML任务中获得最佳性能始终至关重要。
希望对您有帮助。
更新2:
我发现了另一个问题,我也检查并证明是正确的,数字在图像内部的放置也很关键,这对于这种类型的NN是有意义的。一个很好的例子,在MNIST数据集中居中放置的数字 7 和 9 ,如果我们放置新的数字,则图像底部附近会导致较难或易碎的分类用于在图像中心进行分类。我检查了将 7 和 9 移至底部的理论,因此在图像顶部保留了更多位置,结果几乎是 100%准确性。 由于这是一个 spatial 类型的问题,我想通过 CNN 我们可以更有效地消除它。但是,如果MNIST被指定为居中,那会更好,或者我们可以通过编程方式避免出现此问题。