Boundig boxes at the detection are big

qqwweee / keras-yolo3

A Keras implementation of YOLOv3 (Tensorflow backend)

MIT License

7.14k stars 3.45k forks source link

Boundig boxes at the detection are big #737

Open nmargaris opened 3 years ago

nmargaris commented 3 years ago

Hello, i have done training and it s working fine . I am doing tests giving the comand python yolo_video.py --image and for every image the results are good. The only problem is that the detection bounding boxes are bigger than I would like. Is there a way to "zoom" in on their sides towards their center?

pgw928 commented 3 years ago

You can change the size of anchor boxes. Check yolo_anchors.txt out.