As suggested, the FPS using asynchronous mode suppose to be higher than synchronous. However, my model's FPS for asynchronous is lower than synchronous mode. Is there any explanation or solution for this? Besides, I found that the average latency for synchronous mode can be calculated by dividing the duration by count but it is not applicable for the asynchronous mode. Is there any way to calculate the latency in that case? Thx in advance.
Thanks for reaching out.
From my side, the asynchronous is faster than synchronous as below:
Latency measures inference time (in ms) required to process a single input.
Throughput is calculated by dividing the number of inputs that were processed by the processing time.
The formula is:
FPS = Inferred number of images
Processing time in seconds
Check out this documentation for more information.
This thread will no longer be monitored since we have provided the information. If you need any additional information from Intel, please submit a new question.