Performance metrics for new model

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )

http://pjreddie.com/darknet/

Other

21.65k stars 7.96k forks source link

Performance metrics for new model #4778

Open xtream111 opened 4 years ago

xtream111 commented 4 years ago

I trained a new model using custom pictures and I want to know how to get performance metrics for my model. For instance, in the repo there is this information

yolov3-spp.cfg - 60.6% mAP@0.5 - 30 FPS - 141.5 BFlops - 240 MB: yolov3-spp.weights

How do I calculate these numbers.

Thank you

AlexeyAB commented 4 years ago

mAP - ./darknet detector map coco.data yolov3.cfg yolov3.weights
FPS - ./darknet detector demo coco.data yolov3.cfg yolov3.weights -benchmark
MB - just look at the file size yolov3-spp.weights
BFlops - just run any command

xtream111 commented 4 years ago

Is there any way to calculate the inference time?

AlexeyAB commented 4 years ago

inference time = 1000 / FPS