search

endernewton / tf-faster-rcnn

Tensorflow Faster RCNN for Object Detection
https://arxiv.org/pdf/1702.02138.pdf
MIT License
3.65k stars 1.58k forks source link

Non Standard Anchor Shapes #443

Open datNurd opened 5 years ago

datNurd commented 5 years ago

I would like to understand why the anchor sizes are (8,16,24) instead of the standard (128,256,512) recommended/used in the paper ?

ZhuoyaYang commented 5 years ago

You can see generate_anchors.py, base_size=16, 16*(8,16,32) = (128,256,512)