python - 'LinearSVC' 对象没有属性 'classes_'

Question

我有几个图像样本，我想预测这些图像是否包含文本/字符。

当我尝试在这一步运行我的代码时出现错误:

model = cPickle.load(f)
is_text = model.predict(image_samples)

image_samples 是我的样本，模型看起来像这样:

Pipeline(steps=[
    ('hog', HOGFeatures(cells_per_block=(2, 2), orientations=10,
                        pixels_per_cell=(5, 5), size=(20, 20))),
    ('clf', LinearSVC(C=2.0, class_weight=None, dual=True,
                      fit_intercept=True,
                      intercept_scaling=1, loss='l2', max_iter=None,
                      multi_class='ovr', penalty='l2',
                      random_state=None, tol=0.0001, verbose=0))
])

我得到的错误信息是:

File "/home/parallels/Desktop/Python/ImageTextRecognition-master/userimageski.py", line 104, in select_text_among_candidates
is_text = model.predict(self.candidates['flattened'])

File "/usr/local/lib/python2.7/dist-packages/sklearn/utils/metaestimators.py", line 37, in <lambda>
out = lambda *args, **kwargs: self.fn(obj, *args, **kwargs)

File "/usr/local/lib/python2.7/dist-packages/sklearn/pipeline.py", line 180, in predict
return self.steps[-1][-1].predict(Xt)

File "/usr/local/lib/python2.7/dist-packages/sklearn/linear_model/base.py", line 228, in predict
return self.classes_[indices]
AttributeError: 'LinearSVC' object has no attribute 'classes_'

谁能帮帮我？非常感谢

Answer 1

我遇到了同样的问题。这实际上是 sklearn 版本的问题。在终端中运行 python2.7 并检查 >>> import sklearn >>> sklearn.__version__ 如果你有比当前版本更早的版本和截至目前的 0.17.1是最新的，那么你需要升级 sklearn 。只需执行 sudo pip install -U scikit-learn 即可升级并重新运行代码