-
公开(公告)号：US09799098B2
公开(公告)日：2017-10-24
申请号：US12597406
申请日：2008-04-24
申请人： H. Sebastian Seung , Joseph F. Murray , Viren Jain , Srinivas C. Turaga , Moritz Helmstaedter , Winfried Denk
发明人： H. Sebastian Seung , Joseph F. Murray , Viren Jain , Srinivas C. Turaga , Moritz Helmstaedter , Winfried Denk
CPC分类号： G06T5/001 , G06K9/342 , G06K9/4628 , G06T7/11 , G06T2207/20084
摘要： Identifying objects in images is a difficult problem, particularly in cases an original image is noisy or has areas narrow in color or grayscale gradient. A technique employing a convolutional network has been identified to identify objects in such images in an automated and rapid manner. One example embodiment trains a convolutional network including multiple layers of filters. The filters are trained by learning and are arranged in successive layers and produce images having at least a same resolution as an original image. The filters are trained as a function of the original image or a desired image labeling; the image labels of objects identified in the original image are reported and may be used for segmentation. The technique can be applied to images of neural circuitry or electron microscopy, for example. The same technique can also be applied to correction of photographs or videos.
-
公开(公告)号：US20100183217A1
公开(公告)日：2010-07-22
申请号：US12597406
申请日：2008-04-24
申请人： H. Sebastian Seung , Joseph F. Murray , Viren Jain , Srinivas C. Turaga , Moritz Helmstaedter , Winfried Denk
发明人： H. Sebastian Seung , Joseph F. Murray , Viren Jain , Srinivas C. Turaga , Moritz Helmstaedter , Winfried Denk
IPC分类号： G06K9/62
CPC分类号： G06T5/001 , G06K9/342 , G06K9/4628 , G06T7/11 , G06T2207/20084
摘要： Identifying objects in images is a difficult problem, particularly in cases an original image is noisy or has areas narrow in color or grayscale gradient. A technique employing a convolutional network has been identified to identify objects in such images in an automated and rapid manner. One example embodiment trains a convolutional network including multiple layers of filters. The filters are trained by learning and are arranged in successive layers and produce images having at least a same resolution as an original image. The filters are trained as a function of the original image or a desired image labeling; the image labels of objects identified in the original image are reported and may be used for segmentation. The technique can be applied to images of neural circuitry or electron microscopy, for example. The same technique can also be applied to correction of photographs or videos.
摘要翻译： 识别图像中的对象是一个困难的问题，特别是在原始图像嘈杂或具有颜色或灰度梯度窄的区域的情况下。 已经鉴定了采用卷积网络的技术，以自动和快速的方式识别这些图像中的对象。 一个示例性实施例训练包括多层滤波器的卷积网络。 过滤器通过学习进行训练，并且以连续的层布置并产生具有与原始图像至少相同分辨率的图像。 根据原始图像或期望的图像标记来对滤波器进行训练; 报告原始图像中识别的对象的图像标签，并可用于分割。 该技术可以应用于例如神经电路或电子显微镜的图像。 相同的技术也可以应用于照片或视频的校正。
-
公开(公告)号：US20100266175A1
公开(公告)日：2010-10-21
申请号：US12760668
申请日：2010-04-15
IPC分类号： G06T7/00
CPC分类号： G06T7/0081 , G06T7/11 , G06T7/12 , G06T7/162 , G06T7/187 , G06T2207/20081
摘要： Approaches to segmentation or detection of objects and their boundaries in images (or other data sets) do not rely on machine learning approaches that aim to minimize pixel-level agreement between a computer and a human. Optimizing such pixel-level agreement does not, in general, provide the best possible result if boundary detection is a means to the ultimate goal of image segmentation, rather than an end in itself. In some examples, end-to-end learning of image segmentation specifically targets boundary errors with topological consequences, but otherwise does not require the computer to “slavishly” imitate human placement of boundaries. In some examples, this is accomplished by modifying a standard learning procedure such that human boundary tracings are allowed to change during learning, except at locations critical to preserving topology.
摘要翻译： 在图像（或其他数据集）中分割或检测对象及其边界的方法不依赖于旨在最小化计算机与人类之间的像素级协议的机器学习方法。 如果边界检测是图像分割的最终目标的手段，而不是本身的结束，则优化这样的像素级协议通常不提供最佳可能的结果。 在一些例子中，图像分割的端到端学习专门针对具有拓扑结果的边界错误，但是否则不需要计算机“妄图”模拟人类边界的位置。 在一些示例中，这通过修改标准学习过程来实现，使得人学边界追踪在学习期间被允许改变，除了在保持拓扑关键的位置之外。
-
公开(公告)号：US08885926B2
公开(公告)日：2014-11-11
申请号：US12760668
申请日：2010-04-15
CPC分类号： G06T7/0081 , G06T7/11 , G06T7/12 , G06T7/162 , G06T7/187 , G06T2207/20081
摘要： Approaches to segmentation or detection of objects and their boundaries in images (or other data sets) do not rely on machine learning approaches that aim to minimize pixel-level agreement between a computer and a human. Optimizing such pixel-level agreement does not, in general, provide the best possible result if boundary detection is a means to the ultimate goal of image segmentation, rather than an end in itself. In some examples, end-to-end learning of image segmentation specifically targets boundary errors with topological consequences, but otherwise does not require the computer to “slavishly” imitate human placement of boundaries. In some examples, this is accomplished by modifying a standard learning procedure such that human boundary tracings are allowed to change during learning, except at locations critical to preserving topology.
摘要翻译： 在图像（或其他数据集）中分割或检测对象及其边界的方法不依赖于旨在最小化计算机与人类之间的像素级协议的机器学习方法。 如果边界检测是图像分割的最终目标的手段，而不是本身的结束，则优化这样的像素级协议通常不提供最佳可能的结果。 在一些例子中，图像分割的端到端学习专门针对具有拓扑结果的边界错误，但是否则不需要计算机“妄图”模拟人类边界的位置。 在一些示例中，这通过修改标准学习过程来实现，使得人学边界追踪在学习期间被允许改变，除了在保持拓扑关键的位置之外。
-
-
-
搜索结果
4 条记录 (耗时 0.017 秒)
