专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

1. 发明公开

US20240046100A1 Apparatus, method and computer program for decoding neural network parameters and apparatus, method and computer program for encoding neural network parameters using an update model 审中-公开
公开(公告)号：US20240046100A1
公开(公告)日：2024-02-08
申请号：US18380551
申请日：2023-10-16
申请人： Fraunhofer-Gesellschaft zur Foerderung der angewandten Forschung e.V.
发明人： Paul HAASE , Heiner KIRCHHOFFER , Daniel BECKING , Gerhard TECH , Karsten MUELLER , Wojciech SAMEK , Heiko SCHWARZ , Detlev MARPE , Thomas WIEGAND
IPC分类号： G06N3/082
CPC分类号： G06N3/082
摘要： Embodiments according to the invention comprise an apparatus for decoding neural network parameters, which define a neural network. The apparatus may, optionally, be configured to obtain, e.g. to decode, parameters of a base model, e.g. NB, of the neural network which define one or more layers, e.g. base layers, of the neural network. Furthermore, the apparatus is configured to decode an update model, e.g. NU1 to NUK, which defines a modification of one or more layers, e.g. base layers, of the neural network, and the apparatus is configured modify parameters of a base model of the neural network using the update model, in order to obtain an updated model, e.g. designated as “new model” comprising new model layers LNkj. Moreover, the apparatus is configured to evaluate a skip information, e.g. a skip_row_flag and/or a skip_column_flag, indicating whether a sequence, e.g. a row, or a column or a block, of parameters of the update model is zero or not.

2. 发明申请

US20140313290A1 CONCEPT FOR DETERMINING A MEASURE FOR A DISTORTION CHANGE IN A SYNTHESIZED VIEW DUE TO DEPTH MAP MODIFICATIONS 有权
标题翻译：概念确定深度地图修改合成视图中的失真变化的测量方法
公开(公告)号：US20140313290A1
公开(公告)日：2014-10-23
申请号：US14272690
申请日：2014-05-08
申请人： Fraunhofer-Gesellschaft zur Foerderung der angewandten Forschung e.V.
发明人： Gerhard TECH , Karsten MUELLER , Thomas WIEGAND
IPC分类号： H04N13/00
CPC分类号： H04N13/161 , G06T15/205 , H04N13/111 , H04N13/122 , H04N13/128
摘要： An apparatus for determining a measure for a distortion change of a first view synthesized from a second view, caused by a modification of a depth map of the second view from a first state to a second state, is configured—starting from a current synthesis state of the first view corresponding to a synthesis from the second view having the depth map modified to the second state in an already processed portion of the depth map and having the depth map unmodified at the first state in a yet to be processed portion of the depth map—to compute a possible successor synthesis state corresponding to a synthesis of the first view from the second view having the depth map modified to the second state in an already processed portion plus a currently processed portion and having the depth map unmodified at the first state in the yet to be processed portion without the currently processed portion; and to determine a distortion change of a distortion of the current synthesis state of the first view relative to an undistorted version of the first view to a distortion of the possible successor synthesis state of the first view relative to the undistorted version of the first view.
摘要翻译：一种用于确定由第二视图从第一状态到第二状态的深度图的修改引起的从第二视点合成的第一视图的失真变化的测量装置，从当前合成状态开始所述第一视图对应于在所述深度图的已处理部分中具有修改为所述第二状态的所述深度图的第二视图的合成，并且在所述深度的未处理部分中在所述第一状态下未修改所述深度图映射 - 从已经处理部分加上当前处理部分的深度图修改为第二状态的第二视图的第一视图的合成计算可能的后继合成状态，并且在第一状态下未修改深度图在尚未被处理的部分中，没有当前处理的部分; 并且确定第一视图的当前合成状态相对于第一视图的未失真版本的失真的失真变形，以相对于第一视图的未失真版本的第一视图的可能的后继合成状态的失真。

3. 发明公开

US20240046093A1 DECODER, ENCODER, CONTROLLER, METHOD AND COMPUTER PROGRAM FOR UPDATING NEURAL NETWORK PARAMETERS USING NODE INFORMATION 审中-公开
公开(公告)号：US20240046093A1
公开(公告)日：2024-02-08
申请号：US18487885
申请日：2023-10-16
申请人： Fraunhofer-Gesellschaft zur Foerderung der angewandten Forschung e.V.
发明人： Heiner KIRCHHOFFER , Karsten MÜLLER , Paul HAASE , Daniel BECKING , Gerhard TECH , Wojciech SAMEK , Heiko SCHWARZ , Detlev MARPE , Thomas WIEGAND
IPC分类号： G06N3/08
CPC分类号： G06N3/08
摘要： Disclosed is a decoder for decoding parameters of a neural network, configured to obtain a plurality of neural network parameters of the neural network on the basis of an encoded bitstream, to obtain, e.g. to receive; e.g. to extract from an encoded bitstream, a node information describing a node of a parameter update tree, wherein the node information has a parent node identifier, which is, for example, a unique parent node identifier, for example an integer number, a string, and/or a cryptographic hash, and wherein the node information has a parameter update information, e.g. one or more update instructions, for example a difference signal between initial neural network parameters and a newer version thereof, e.g. corresponding to a child node of the update tree, and to derive one or more neural network parameters using parameter information of a parent node identified by the parent node identifier and using the parameter update information.

4. 发明申请

US20180027258A1 3D-VIDEO CODEC SUPPORTING INTER-COMPONENT PREDICTION 审中-公开
公开(公告)号：US20180027258A1
公开(公告)日：2018-01-25
申请号：US15669063
申请日：2017-08-04
申请人： Fraunhofer-Gesellschaft zur Foerderung der angewandten Forschung e.V.
发明人： Gerhard TECH , Karsten SUEHRING , Detlev MARPE , Thomas WIEGAND
IPC分类号： H04N19/597 , H04N19/54 , H04N19/58 , H04N19/543
CPC分类号： H04N19/597 , H04N19/105 , H04N19/157 , H04N19/187 , H04N19/54 , H04N19/543 , H04N19/58 , H04N19/70
摘要： 3D video coding/decoding supporting inter-component prediction is made operatable with respect to a broader range of settings with respect to accompanying individual views with depth and deactivation/enablement of inter-component prediction, by using one or more flags in the data stream to control the enablement/disablement of the availability of inter-component prediction for coding units of a current picture.

5. 发明申请

US20170332101A1 3D VIDEO DATA STREAM COMPRISING A VALUE TABLE ASSOCIATING REFERENCE VIEWS AND CAMERA PARAMETER IN TEMPORAL UNIT SCOPE AND BEYOND TEMPORAL UNIT SCOPE PARAMETER SETS AND AN ENCODER FOR ENCODING AND A DECODER FOR DECODING THE 3D VIDEO DATA STREAM 审中-公开
公开(公告)号：US20170332101A1
公开(公告)日：2017-11-16
申请号：US15668939
申请日：2017-08-04
申请人： Fraunhofer-Gesellschaft zur Foerderung der angewandten Forschung e.V.
发明人： Gerhard TECH , Karsten SUEHRING , Detlev MARPE , Thomas WIEGAND
IPC分类号： H04N19/597 , H04L29/06 , H04N19/463
CPC分类号： H04N19/597 , H04L65/607 , H04N19/463 , H04N19/70
摘要： A 3D video data stream including a set of coded views coded in the 3D video data stream in temporal units is shown. The 3D video data stream further includes a parameter set including a first table having an entry for each coded view which indicates by indexing for the respective coded view a number of reference views for which a relative camera parameter is present in the data stream, wherein a scope of the parameter set is beyond the scope of the temporal units.

6. 发明申请

US20140247872A1 EFFECTIVE WEDGELET PARTITION CODING 审中-公开
标题翻译：有效的WEDGELET分区编码
公开(公告)号：US20140247872A1
公开(公告)日：2014-09-04
申请号：US14273609
申请日：2014-05-09
申请人： Fraunhofer-Gesellschaft zur Foerderung der angewandten Forschung e.V.
发明人： Philipp MERKLE , Christian BARTNIK , Haricharan LAKSHMAN , Detlev MARPE , Karsten MUELLER , Thomas WIEGAND , Gerhard TECH
IPC分类号： H04N19/597 , H04N19/167 , H04N19/124 , H04N19/44 , H04N19/593 , H04N19/174
CPC分类号： H04N19/105 , H04N19/119 , H04N19/124 , H04N19/157 , H04N19/167 , H04N19/174 , H04N19/176 , H04N19/44 , H04N19/46 , H04N19/593 , H04N19/597 , H04N19/70 , H04N19/82 , H04N19/96
摘要： Wedgelet separation lines of neighboring blocks are predicted from each other by aligning the wedgelet separation lines of such neighboring blocks to each other, thereby reducing the side information coding rate that may be used.
摘要翻译：通过将这些相邻块的楔形间隔线彼此对准来预测相邻块的楔形线分隔线，从而减少可以使用的侧信息编码率。

7. 发明申请

US20180063550A1 MULTI-VIEW VIDEO CODEC SUPPORTING RESIDUAL PREDICTION 审中-公开
公开(公告)号：US20180063550A1
公开(公告)日：2018-03-01
申请号：US15668996
申请日：2017-08-04
申请人： Fraunhofer-Gesellschaft zur Foerderung der angewandten Forschung e.V.
发明人： Gerhard TECH , Karsten SUEHRING , Detlev MARPE , Thomas WIEGAND
IPC分类号： H04N19/597 , H04N19/42 , H04N19/105 , H04N19/573 , H04N19/70 , H04N19/65
CPC分类号： H04N19/597 , H04N19/103 , H04N19/105 , H04N19/176 , H04N19/42 , H04N19/573 , H04N19/61 , H04N19/65 , H04N19/70
摘要： Multi-view video coding/decoding supporting residual prediction involving a reference picture of different view and timestamp compared to a currently coded/decoded picture is improved in terms of robustness and/or processing cost by avoiding checking any picture availability or non-availability in a decoded picture buffer.

8. 发明申请

US20140247871A1 ADAPTIVE PARTITION CODING 有权
标题翻译：自适应分区编码
公开(公告)号：US20140247871A1
公开(公告)日：2014-09-04
申请号：US14273601
申请日：2014-05-09
申请人： Fraunhofer-Gesellschaft zur Foerderung der angewandten Forschung e.V.
发明人： Philipp MERKLE , Christian BARTNIK , Haricharan LAKSHMAN , Detlev MARPE , Karsten MUELLER , Thomas WIEGAND , Gerhard TECH
IPC分类号： H04N19/119 , H04N19/593 , H04N19/44
CPC分类号： H04N19/176 , H04N19/103 , H04N19/105 , H04N19/119 , H04N19/14 , H04N19/157 , H04N19/196 , H04N19/23 , H04N19/44 , H04N19/46 , H04N19/593 , H04N19/597 , H04N19/70 , H04N19/82 , H04N19/96
摘要： Although wedgelet-based partitioning seems to represent a better tradeoff between side information rate on the one hand and achievable variety in partitioning possibilities on the other hand, compared to contour partitioning, the ability to alleviate the constraints of the partitioning to the extent that the partitions have to be wedgelet partitions, enables applying relatively uncomplex statistical analysis onto overlaid spatially sampled texture information in order to derive a good predictor for the bi-segmentation in a depth/disparity map. Thus, in accordance with a first aspect it is exactly the increase of the freedom which alleviates the signaling overhead provided that co-located texture information in form of a picture is present. Another aspect pertains the possibility to save side information rate involved with signaling a respective coding mode supporting irregular partitioning.
摘要翻译：尽管基于楔形的划分似乎代表了侧面信息速率与另一方面可实现的分区可能性之间的更好的权衡，与轮廓划分相比，能够将分区的约束减轻到分区的程度必须是楔形分割，使得能够对覆盖的空间采样纹理信息应用相对不复杂的统计分析，以便为深度/视差图中的双分割导出良好的预测器。因此，根据第一方面，如果存在图像形式的共位纹理信息，则恰好是减轻信令开销的自由度的增加。另一方面涉及节省支持不规则分区的相应编码模式的信令所涉及的侧信息速率的可能性。

9. 发明申请

US20140341290A1 EFFECTIVE WEDGELET PARTITION CODING USING SPATIAL PREDICTION 有权
标题翻译：使用空间预测的有效的卫星分割编码
公开(公告)号：US20140341290A1
公开(公告)日：2014-11-20
申请号：US14273600
申请日：2014-05-09
申请人： Fraunhofer-Gesellschaft zur Foerderung der angewandten Forschung e.V.
发明人： Philipp MERKLE , Christian BARTNIK , Haricharan LAKSHMAN , Detlev MARPE , Karsten MUELLER , Thomas WIEGAND , Gerhard TECH
IPC分类号： H04N19/51 , H04N19/583
CPC分类号： H04N19/51 , H04N19/105 , H04N19/119 , H04N19/14 , H04N19/157 , H04N19/176 , H04N19/196 , H04N19/46 , H04N19/52 , H04N19/593 , H04N19/597 , H04N19/70 , H04N19/82 , H04N19/96
摘要： In accordance with a first aspect, the intra prediction direction of a neighboring, intra-predicted block is used in order to predict the extension direction of the wedgelet separation line of a current block, thereby reducing the side information rate necessitated in order to convey the partitioning information. In accordance with a second aspect, the idea is that previously reconstructed samples, i.e. reconstructed values of blocks preceding the current block in accordance with the coding/decoding order allow for at least a prediction of a correct placement of a starting point of the wedgelet separation line, namely by placing the starting point of the wedgelet separation line at a position of a maximum change between consecutive ones of a sequence of reconstructed values of samples of a line of samples extending adjacent to the current block along a circumference thereof. Both aspects may be used individually or in combination.
摘要翻译：根据第一方面，使用相邻的帧内预测块的帧内预测方向来预测当前块的楔形分离线的延伸方向，从而减少为了传送所述帧预测块所需的侧信息速率分区信息。根据第二方面，该想法是，先前重建的样本，即根据编码/解码顺序在当前块之前的块的重建值允许至少预测楔形分离的起点的正确放置即通过将楔形分离线的起始点放置在沿着其圆周与当前块相邻延伸的一系列样本的样本重建值的序列中的连续的重复值之间的最大变化的位置。这两个方面可以单独使用或组合使用。

10. 发明申请

US20140247867A1 EFFECTIVE PREDICTION USING PARTITION CODING 有权
标题翻译：使用分区编码的有效预测
公开(公告)号：US20140247867A1
公开(公告)日：2014-09-04
申请号：US14273603
申请日：2014-05-09
申请人： Fraunhofer-Gesellschaft zur Foerderung der angewandten Forschung e.V.
发明人： Philipp MERKLE , Christian BARTNIK , Haricharan LAKSHMAN , Detlev MARPE , Karsten MUELLER , Thomas WIEGAND , Gerhard TECH
IPC分类号： H04N19/597 , H04N19/126
CPC分类号： H04N19/176 , H04N19/105 , H04N19/119 , H04N19/126 , H04N19/14 , H04N19/157 , H04N19/196 , H04N19/46 , H04N19/593 , H04N19/597 , H04N19/61 , H04N19/70 , H04N19/82 , H04N19/96
摘要： The way of predicting a current block by assigning constant partition values to the partitions of a bi-partitioning of a block is quite effective, especially in case of coding sample arrays such as depth/disparity maps where the content of these sample arrays is mostly composed of plateaus or simple connected regions of similar value separated from each other by steep edges. The transmission of such constant partition values would, however, still need a considerable amount of side information which should be avoided. This side information rate may be further reduced if mean values of values of neighboring samples associated or adjoining the respective partitions are used as predictors for the constant partition values.
摘要翻译：通过将恒定分区值分配给块的双分区的分区来预测当前块的方法是非常有效的，特别是在编码诸如深度/视差图的样本阵列的情况下，其中这些样本阵列的内容主要组成通过陡峭的边缘彼此相隔的平台或简单连接区域。然而，这种恒定分区值的传输仍然需要应该避免的相当多的辅助信息。如果将相关分区相邻或邻接的相邻样本的值的平均值用作恒定分区值的预测值，则可以进一步减少该侧信息速率。

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式