Two channels float 32 images pipeline

[raphenri]

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) Jetson AGX orin
• DeepStream Version 7
• JetPack Version (valid for Jetson only) 6.0
• TensorRT Version 8.6
**• Cuda 12.2
• Issue Type( questions, new requirements, bugs) questions

Hi,

I have trained a UNET segmentation model with float32 images of two channels (RG).
I would like to use deepstream to run a pipeline and run inference.

I think I can use nvdspreprocess to put my data in the right format (float32 2 channels) but I am not sure. Do you have an example of nvdspreprocess usage ?
Also how can I use nvinfer after nvdspreprocess in that context, since nvinfer expects RGB or GRAY images and int8 only ? My model expect float32-two channels images so is there a way to configure nvinfer in such a way ?

[fanzh]

please refer to this topic.
if using nvdspreprocess plugin, nvinfer will receive preprocessed data and does not need to do preprocessing again. nvinfer will send the preprocessed tensors to do inference directly. nvinfer plugin and low-level are openousrce. from the TensorRT interface enqueueBuffer, this color format is not need.
especially please set input-tensor-meta of nvinfer to 1. please refer to sample deepstream-preprocess-test.

[yingliu]

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks