How to get better mAP in xnor net training

AlexeyAB / darknet

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

http://pjreddie.com/darknet/

Other

21.68k stars 7.96k forks source link

How to get better mAP in xnor net training #2556

Open dudqls1994 opened 5 years ago

dudqls1994 commented 5 years ago

Hello, i 'm training xnor-net used voc data training DB : voc 12 + 7 test DB : voc 7 cfg : tiny-yolo-voc_xnor.cfg weight : yolov2-tiny-voc.weights lr : 0.001

I get only mAP 20% .... Any one help me how to get better mAP in Xnor net?

AlexeyAB commented 5 years ago

@dudqls1994 Hi,

You should use pre-trained XNOR weights file that is trained for XNOR Classification model on ImageNet ~1.3M images, like here: https://github.com/AlexeyAB/darknet/issues/2365#issuecomment-462923756