# Configure ML plugins

The ML plugins facilitate the preprocessing, inferencing, and postprocessing of the machine learning models.

## Related information

- [Configure display, camera, encode and decode plugins](https://docs.qualcomm.com/doc/80-70029-50/topic/display-plugins.html)
- [Configure audio plugins](https://docs.qualcomm.com/doc/80-70029-50/topic/audio-plugins.html)

- [qtimlvconverter](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimlvconverter.html)
The qtimlvconverter plugin transforms the incoming video buffers into neural-network tensors while performing necessary format conversion and resizing in the process. To achieve these operations, the plugin uses the GPU hardware and ION/DMA allocated buffers.
- [qtimlaconverter](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimlaconverter.html)
The qtimlaconverter plugin processes the incoming audio waveform data into ML tensors. The ML models such as the audio classification model process these tensors for inferencing.
- [qtibatch](https://docs.qualcomm.com/doc/80-70029-50/topic/qtibatch.html)
The qtibatch plugin uses frame aggregation techniques to group several audio/video frames into one buffer for preprocessing. Batching can be done on a single stream (in which case the batch size is 1) or in many parallel streams.
- [qtimlsnpe](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimlsnpe.html)
The qtimlsnpe plugin shows the Qualcomm^®^ Neural Processing SDK capabilities (load and run the models).
- [qtimltflite](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimltflite.html)
The qtimltflite plugin shows the LiteRT capabilities (load and run the LiteRT models) as a GStreamer plugin.
- [qtimlqnn](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimlqnn.html)
The qtimlqnn plugin shows the Qualcomm^®^ AI Engine direct SDK capabilities (load and execute the Qualcomm Neural Network models) as a GStreamer plugin.
- [qtimlonnx](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimlonnx.html)
The mlonnx plugin provides ONNX (Open Neural Network Exchange) model inference capabilities within GStreamer pipelines. It leverages ONNX Runtime with support for multiple execution providers including CPU and Qualcomm QNN (Qualcomm Neural Network) hardware acceleration.
- [qtimlpostprocess](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimlpostprocess.html)
The qtimlpostprocess is a customizable plugin that provides a library interface to postprocess the tensor output of the inference plugins. The postprocessing library is solely responsible to parse the tensor and generate a list of predicted output modes. The plugin manages the module execution, output generation (ML metadata or image mask), batching, ML staging, and other related tasks.
- [qtimlvclassification](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimlvclassification.html)
The qtimlvclassification plugin processes output tensors of an image classification model from the ML inference plugin (such as qtimltflite, qtimlsnpe, and qtimlqnn) into a result of predictions.
- [qtimlaclassification](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimlaclassification.html)
The qtimlaclassification plugin processes the output tensors of an audio classification model from the ML inference plugin (such as qtimltflite) into a result of predictions.
- [qtimlvdetection](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimlvdetection.html)
The qtimlvdetection plugin processes output tensors of an object detection model from the ML inference plugin (such as qtimltflite, qtimlsnpe, and qtimlqnn) into result of predictions.
- [qtimlvsegmentation](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimlvsegmentation.html)
The qtimlvsegmentation plugin processes output tensors of an image segmentation/depth estimation model from the ML inference plugin (such as qtimltflite, qtimlsnpe, and qtimlqnn) into result of predictions.
- [qtimlvpose](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimlvpose.html)
The qtimlvpose plugin processes output tensors of a pose estimation model from the ML inference plugin (such as qtimltflite, qtimlsnpe, and qtimlqnn) into result of predictions.
- [qtimlvsuperresolution](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimlvsuperresolution.html)
The qtimlvsuperresolution plugin processes output tensors of an image super resolution model from the ML inference plugin (such as qtimltflite or qtimlsnpe).
- [qtivoverlay](https://docs.qualcomm.com/doc/80-70029-50/topic/qtioverlay.html)
The qtivoverlay plugin is a hardware accelerated in-place image draw and blit plugin for drawing overlays on top of the YUV images.
- [qtimetamux](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimetamux.html)
The qtimetamux plugin uses frame matching techniques to associate or attach ML string based postprocessing results (output from the postprocessing plugin) or CV information to original frame such as [GstMeta](https://gstreamer.freedesktop.org/documentation/gstreamer/gstmeta.html#GstMeta).
- [qtimldemux](https://docs.qualcomm.com/doc/80-70029-50/topic/qtimldemux.html)
The qtimldemux element splits batched (such as first tensor dimension is greater and 1) tensors (GstMemory blocks) from a single input GstBuffer into separate GstBuffers containing an unbatched tensors (GstMemory blocks).
- [qtirtspbin](https://docs.qualcomm.com/doc/80-70029-50/topic/qtirtspbin.html)
The qtirtspbin plugin transmits the video or data streams using RTSP.
- [qtismartvencbin](https://docs.qualcomm.com/doc/80-70029-50/topic/qtismartvencbin.html)
The qtismartvencbin plugin uses machine learning and computer vision to reduce the video bandwidth. It can be used for surveillance, where only the small image areas are in motion. The bitrate reduction can be configured according to the quality requirements.

Last Published: Jun 03, 2026

[Previous Topic
pulsesink](https://docs.qualcomm.com/bundle/publicresource/80-70029-50/topics/pulsesink.md) [Next Topic
qtimlvconverter](https://docs.qualcomm.com/bundle/publicresource/80-70029-50/topics/qtimlvconverter.md)