关于of 后使用冠词的总结
关于of后使用冠词的总结：
一直想弄清楚，结果搜索了几次，都没有找到一个完整的解释。干脆自己总结了，有不妥之处，请指正！
Ⅰ. 一般在of后的名词不可数名词 要有“the”，泛指和专有名词除外（多用泛指）：
Segmentation is a by-productof recognition:
the paper of Russell et al
The statistics of the data,
consisting of the many variantsof spectral factorization
that can describe segments of complex shape and appearance这里of后没有冠词
the model of Fig. 1
the shape ofthe segment
the set ofDirichlet Process Gaussian mixture model (DP-GMM)
the level of confidence
the model of (5)
the SPMM of Section 2.1.3
the properties ofthe SPMM
an object ofcomplex shape (the stone arch),
the nonparametric approximationof x.
from averaging patches of different color but similar gradient pattern
the similarity of segmentation accuracy
a certain amount of supervision,
a certain amount of supervision is
the bottom three rows of Table 2
a drop of 0.16 in accuracy.
the use of human supervision
The lack of supervision
Ⅱ. 一般of后的单数可数名词，要有不定冠词a 或 定冠词 the， 除非是用了“each, every, that, my，both…”等，如：
the shape and appearance of each segment
the components ofthe object
the appearance of an image
the segments of each image and the correspondence between such segments
Plate diagram ofourgenerative model
the number of segment？？？？
the nth observation of image m????（有问题？）
a building block of a larger model
The choice of a Dirichlet distribution
The advantage ofour model is
Ⅲ. 如果是复数家-s，前面可以不要定冠词“the”除非特别指定。如：
the popular “bag of visual words” model
a collection of images
a unified view of both tasks
a small dictionary of “textons,”
where statistics of words are used to
the visual appearance of different segments in images.
independently of the visual words
the shape of clusters
segments of complex shapes
the use of superpixels
the total number of pixels
all the images offaces,
the features that relate segments are restricted to size,shape, and average color ofthe segments.
Ⅳ. 特殊的一类固定搭配of后接单数，如：
this type of model is used
the variables of interest,
Ⅴ. 在of后接动名词，如
the idea of carrying out category learning for recognition and segmentation jointly.
The performance of fitting a GMM is of the lowest quality
The appearance of recurring elements
复合示例句子：
Our model represents each image as a composition of segments, where a segment could correspond to a whole object (e.g., a cow) or to a partof an object (e.g., a leg), to a patch of a distinctive texture, or to a “nonsense” homogeneous region in the background.
The more challenging problems of simultaneous segmentation and recognition of an entire image collection,…
The shape and appearance of a segment across a collection of imagesof objects…
以上是从一篇paper中摘录下来的，《Unsupervised Learning of Categorical Segments in Image Collections》PAMI2012
I couldn'tthink of the name of that man anyhow.
我怎么也想不出那人的名字来了。
Such recurring segments may bethought of as the “parts” of corresponding objects that appear multiple times in the image collection.(考虑成…)