Closed chakloong883 closed 2 years ago
Every time you call "execute()" on a TensorRT engine, it would process 1 batch of sample (e.g. 1 image).
In other words, if you don't call "execute()" continuously, the engine will not continue to do inference or detection.
Take trt_yolo.py for example, I have wrapped the "execute()" call inside TrtYOLO's "detect()" function. I call this "detect()" in a while loop so it does detection continuously.
If you understand the source code, you could just put this "detect()" call at the proper place. There's no need for a pause function.
Every time you call "execute()" on a TensorRT engine, it would process 1 batch of sample (e.g. 1 image).
In other words, if you don't call "execute()" continuously, the engine will not continue to do inference or detection.
Take trt_yolo.py for example, I have wrapped the "execute()" call inside TrtYOLO's "detect()" function. I call this "detect()" in a while loop so it does detection continuously.
If you understand the source code, you could just put this "detect()" call at the proper place. There's no need for a pause function.
Thank you so much! I learned a lot, you can close the issue.
I want to develop on qt and design a pause function, I want to know how to stop tensorrt's inference engine to avoid errors when continuing detection. Looking forward to your answer