Skip to content

commands.md

Latest commit

 

History

History
56 lines (40 loc) · 1.31 KB

commands.md

File metadata and controls

56 lines (40 loc) · 1.31 KB

Useful comands:

to get yolo running on webcam:

  • ./darknet detector demo cfg/coco.data cfg/yolo.cfg yolo.weights
  • ./darknet detector demo cfg/voc.data cfg/tiny-yolo-voc.cfg tiny-yolo-voc.weights

Start labeling tool:

  • python main.py

reshape/rename images:

mogrify -resize 320x320 *.jpg
rename 's/.jpg$/.JPEG/' *.jpg

Testing

run on webcam

./darknet detector demo cfg/obj.data cfg/yolo-obj-webCam.cfg TrainingTinyTiny/yolo-obj-1200.weights -c 2

-c 2 tag specifies which webcam to use

run on image

./darknet detector test cfg/obj.data cfg/yolo-obj.cfg  backup/yolo-obj_100.weights data/obj/IMAG3635.jpg -thresh 0.45

-thresh 20 determines the threshold

train

./darknet detector train cfg/obj.data cfg/yolo-obj.cfg tiny-yolo-voc.weights

Working on lab PC

ssh -p 58022 w392@atcremers64.informatik.tu-muenchen.de
Enter PW

cd /usr/prakt/w392/BraccioPlayMemory/darknet

nvidia-smi

./darknet detector train cfg/obj.data cfg/yolo-obj.cfg tiny-yolo-voc.weights

Working with Google Cloud Service

(If instance does not restart with GPU, change region to eaurope-west1-d) Create ssh key: add user name Add ssh key to instance (see Google PDF) Install gcloud tool or use Google Cloud Shell Connect with gcloud compute ssh instance-2 --zone europe-west1-d