**• Hardware Platform: RTXA4000
**• DeepStream Version: 6.0
**• TensorRT Version: 8.0.1
**• NVIDIA GPU Driver Version: 470.94
**• Issue Type: Question regarding nvinfer, nvdsosd and setting interval to nonzero.
When using nvinfer with a configuration file containing interval=0
inference is ran on every frame and the bounding boxes are then drawn on every frame. However when we set interval=1
to apply inference on every other frame, no bounding boxes are drawn on the skipped frames, so you will see blinking boxes and labels. How can we render the previous bounding boxes instead? Is there a way to do this besides using a tracking plugin? Or a simple way to store the meta-data generated from nvinfer for future frames that are not being processed by nvinfer?
nvdsosd-transparency-demo-main.zip (30.2 KB)
Instructions
1- Build container
./docker_build.sh
2 - Run container
./docker_run.sh
3 - Build application
mkdir appsink_nvmm/build
cd appsink_nvmm/build
cmake ..
make
4 - Run application
./appsink_nvmm
5 - Skip Inference on every other frame
Change line 74 in config_infer_primary.txt
vim /opt/nvidia/deepstream/deepstream-6.0/samples/configs/deepstream-app/config_infer_primary.txt
interval=1
Re-run the application