# qtimldemux Source: [https://docs.qualcomm.com/doc/80-70022-50SC/topic/qtimldemux.html](https://docs.qualcomm.com/doc/80-70022-50SC/topic/qtimldemux.html) qtimldemux 元素将单个输入 GstBuffer 中的批处理（例如第一个张量维数更大且 1）张量（GstMemory 块）拆分为包含未批处理张量（GstMemory 块）的单独 GstBuffer。 Figure : qtimldemux 架构 Layer_1 Sheet.2 Sheet.3 Sheet.599 qtimldemux qtimldemux Sheet.601 neural-network/tensors dimensions=<<4,50,80,80,85>, <4,300,30... neural-network/tensorsdimensions=<<4,50,80,80,85>, <4,300,300,85> ...<4,500,85>>,type=UINT8 Sheet.607 后处理插件 后处理插件 Sheet.609 模块 模块 Sheet.613 后处理插件 后处理插件 Sheet.615 模块 模块 Sheet.619 后处理插件 后处理插件 Sheet.621 模块 模块 Sheet.624 每个张量的第一个维度表示其批量大小 每个张量的第一个维度表示其批量大小 Sheet.627 Sheet.628 Sheet.633 Sheet.634 Sheet.635 Sheet.636 Sheet.637 Sheet.658 Sheet.301 GstBuffer GstBuffer Sheet.303 Sheet.304 Sheet.305 + GstMemory 0: 张量 0 + GstMemory 1: 张量 1… + GstMemory ‘N-1’: 张... + GstMemory 0:张量 0+ GstMemory 1:张量+ GstMemory ‘N-1’:张量N-1’+ GstMemory ‘N’:张量N’ Sheet.646 + GstMLTensorMeta 0: type: UINT8 n_dimensions: 5 dimensions[ ... + GstMLTensorMeta 0:· type: UINT8· n_dimensions: 5· dimensions[ ]: {1, 50, 80, 80, 85}+ GstMLTensorMeta 1:· type: UINT8· n_dimensions: 4· dimensions[ ]: {1, 300, 300, 85}…+ GstMLTensorMeta ‘N-1’:· type: UINT8· n_dimensions: 2· dimensions[ ]: {1, 100}+ GstMLTensorMeta ‘N’:· type: UINT8· n_dimensions: 3· dimensions[ ]: {1, 500, 85} Sheet.662 Sheet.657 Sheet.630 Sheet.631 Sheet.659 Sheet.650 GstBuffer GstBuffer Sheet.651 Sheet.652 Sheet.653 + GstMemory 0: 张量 0 + GstMemory 1: 张量 1… + GstMemory ‘N-1’: 张... + GstMemory 0:张量 0+ GstMemory 1:张量+ GstMemory ‘N-1’:张量N-1’+ GstMemory ‘N’:张量N’ Sheet.654 + GstMLTensorMeta 0: type: UINT8 n_dimensions: 5 dimensions[ ... + GstMLTensorMeta 0:· type: UINT8· n_dimensions: 5· dimensions[ ]: {4, 50, 80, 80, 85}+ GstMLTensorMeta 1:· type: UINT8· n_dimensions: 4· dimensions[ ]: {4, 300, 300, 85}…+ GstMLTensorMeta ‘N-1’:· type: UINT8· n_dimensions: 2· dimensions[ ]: {4, 100}+ GstMLTensorMeta ‘N’:· type: UINT8· n_dimensions: 3· dimensions[ ]: {4, 500, 85} Sheet.596 ML 推理插件 ML 推理插件 Sheet.673 neural-network/tensors dimensions=<<4,50,80,80,85>, <4,300,30... neural-network/tensorsdimensions=<<4,50,80,80,85>,<4,300,300,85> … <4,500,85>>, type=UINT8 Sheet.674 neural-network/tensors dimensions=<<4,50,80,80,85>, <4,300,30... neural-network/tensorsdimensions=<<4,50,80,80,85>,<4,300,300,85> … <4,500,85>>, type=UINT8 Sheet.675 neural-network/tensors dimensions=<<4,50,80,80,85>, <4,300,30... neural-network/tensorsdimensions=<<4,50,80,80,85>,<4,300,300,85> … <4,500,85>>, type=UINT8 Sheet.676 Sheet.640 Qualcomm Qualcomm Sheet.641 Sheet.643 开源 开源 Sheet.644 ## 继承链 [GObject](https://docs.gtk.org/gobject/) → [GstObject](https://gstreamer.freedesktop.org/documentation/gstreamer/gstobject.html?gi-language=c) → [GstElement](https://gstreamer.freedesktop.org/documentation/gstreamer/gstelement.html?gi-language=c) → GstMLDemux 下表提供了有关 qtimldemux 的端口模板和元素属性的信息。有关用例信息，请参见[图像分类](https://docs.qualcomm.com/doc/80-70022-50SC/topic/gst-ai-classification.html)。 ## 端口配置 | 端口名称 | 功能 | 功能 | 功能 | | --- | --- | --- | --- | | SINK 模板：'sink'


| neural-network/tensors | type： | { (string)INT8, (string)UINT8, (string)INT32, (string)UINT32, (string)FLOAT16, (string)FLOAT32 } | | SRC模板：'src\_%u'


| neural-network/tensors | type： | { (string)INT8, (string)UINT8, (string)INT32, (string)UINT32, (string)FLOAT16, (string)FLOAT32 } | ## 元素配置 Table : qtimlDemux 元素属性 | 属性 | 说明 | | --- | --- | | name | 对象名称


| | parent | 对象的父级


| **Parent Topic:** [配置 ML 插件](https://docs.qualcomm.com/doc/80-70022-50SC/topic/inferencing-plugins.html) Last Published: Nov 05, 2025 [Previous Topic qtimetamux](https://docs.qualcomm.com/bundle/publicresource/80-70022-50SC/topics/qtimetamux.md) [Next Topic qtirtspbin](https://docs.qualcomm.com/bundle/publicresource/80-70022-50SC/topics/qtirtspbin.md)