Helipad Recognition and Vision based Distance Measurement

Recognition with yolo v3

1. Install darknet

2. Follow the steps here

3. Download yolo v3 weights

4. Train

mkdir backup
darknet detector train custom.data yolov3.cfg yolov3.weights -clear 1

5. Infer

darknet detector test custom.data yolov3.cfg backup/yolov3_final.weights ../img/helipad9.jpg -thresh 0.9

6. Try colab notebooks

Train

Predict

Vision based Distance Measurement

[TODO]

Tips

• Convert all png's to jpg: for i in *.png; do convert $i $(basename $i png)jpg ; done

• -clear 1 argument of darknet discards previously trained batch count

• Due to a bug in darknet if you are working in another directory than darknet's, make a link to data/labels directory.