Nvinfer During inference, does TensorRT modify the input image to the resolution required by the model

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) GPU
• DeepStream Version 6.3-docker
• JetPack Version (valid for Jetson only)
• TensorRT Version 8.5
• NVIDIA GPU Driver Version (valid for GPU only) 550.142
When I input multiple sources, my model is trained with 640*640 resolution, and when I use self.streammux.set_property(‘width’, 1920)
self.streammux.set_property(‘height’, 1080) Does it automatically scale to 640 by 640 when given to nvinfer

3

nvinfer will scale the image to 640x640 internally.

4

If my model was trained on 1280 by 1280 images, how can I adjust it so that nvifer receives images at this resolution without using it

5

This is batch resolution. It has nothing to do with the model’s input. Please ignore it.

gst-nvinfer will generate proper tensor input (scaling, normalization, conversion,…) inside the plugin. Please make sure you have configured the gst-nvinfer plugin correctly.