rtmdet video_demo issue

[k-engus]

Hello, I am a student learning object detection and I have a few questions. I trained the rtmdet-s model with custom data and ran it through video_demo. When I used the --show option, the video GUI is extremely slow with 4-6 tasks/s occurring. However, the converted output video shows inference results at normal speed. Is this normal? Previously, when I ran videos with custom YOLO, detection was performed at normal speed, but rtmdet it's too slow