Closed ghost closed 4 years ago
Please try detection_infer_speed.py
to obtain the inference time of your model, and see the theoritical time needed to go through the whole dataset.
Please try
detection_infer_speed.py
to obtain the inference time of your model, and see the theoritical time needed to go through the whole dataset.
Thanks a lot
I tried to regenerate results of TridentNet with tridentnet_r101v2c4_c5_1x.py and tridentnet_r101v2c4_c5_fastapprox_1x.py config files over COCO dataset but it take a few hours for RTX2060 gpu. I just didnt understand what is the time consuming part for test evaluation, why it takes too long? I wonder if I have a problem.