Changing visualization (bounding box) of the result of DetectNet

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

Hello everyone,

I am running the DetectNet application https://github.com/dusty-nv/jetson-inference
with SSD-Mobilenet-v2 model. I would like to modify the code little bit. As you know, DetectNet uses “bounding box” for identifying the detection on the display.

The related function is

bool detectNet::Overlay(float* input, float* output, uint32_t width, uint32_t height, Detection* detections, uint32_t numDetections, uint32_t flags )

in detectNet.cpp. It is then linked it to cudaDetectionOverlay function, in detectNet.cu.

The result shows like:

1920×1080 443 KB

I would like to have only border in my bounding box; instead of covering the detectnet object with color (based on the detected class). Something like that:

There are several options when you run the program with console commands, such as: – output_cvg and – output_bbox. However, I couldn’t find how to set an argument according to my need.

Do you have any idea how I can achieve that? only way it to modify “gpuDetectionOverlay” function ?

Thank you

2

Hi,

Jetson_inference overlay the bounding box to the source image.
You will need to update the function for the rectangle bounding box.

Thanks.

3

Hi a.kizilay,

Did you manage to do this? I’m going to dive deeper into the gpuDetectionOverlay but could use some guidelines.