Open zdj97 opened 1 week ago
You will have to create the teamName_track1_latency.json on your own. You can use limmits_latency_eval.py. as reference. The code generates the {time_to_first_frame, time_to_last_frame, num_frames} fields by running inference with a smaller reference model. You'll have to run inference using the proposed model and calculate the values as done in limmits_latency_eval.py.
You will have to create the teamName_track1_latency.json on your own. You can use limmits_latency_eval.py. as reference. The code generates the {time_to_first_frame, time_to_last_frame, num_frames} fields by running inference with a smaller reference model. You'll have to run inference using the proposed model and calculate the values as done in limmits_latency_eval.py.
So the difference between teamName_track1_latency.json and teamName_track1_latency_normaliser.json is that, teamName_track1_latency_normaliser.json is created by limmits_latency_eval.py, but teamName_track1_latency.json is created by ourself using same logic according to limmits_latency_eval.py? Thanks
Yes.
For track1, does the content of teamName_track1_latency_normaliser.json and teamName_track1_latency.json is the same? I understand the content of teamName_track1_latency_normaliser.json consist of {time_to_first_frame, time_to_last_frame, num_frames}, and i have no idea of the content of teamName_track1_latency.json as it is not introduced in limmits_latency_eval.py.