Filtering for a single object

[bloodymeli]

Hi,

I was wondering if you can explain how to tweak the code for the following use-case: Instead of detecting multiple objects at multiple scales, I'm interested in detecting only the most dominant object of a specific class (in my case, a person). How can I accelerate the code, and break once I discover the largest person in the frame? I plan to use the pre-trained model, so https://github.com/ultralytics/yolov3/wiki/Train-Custom-Data

was not that useful for me.

Thanks!

[barresoft]

Maybe increment confidence using parameters --confidence and --nms_thresh when call detect.py ....