GiraffeDet

This is an Unofficial implementation for "GiraffeDet: A Heavy-Neck Paradigm for Object Detection" with Tensorflow 1.1x

This code is based on my impletment of yolof
I am still running this code in colab and I am not sure whether there needs any conv op(like 3x3) in gfpn, If you add these ops it may costs lots of Display memory. With these ops you can rnu 8 batch with a Tesla K80(free mem about 15G) with resolution of 384x384. Besides it's hard to converge without pretrain backbone, maybe you can switch to * resnet.
Any question are welcomed.Contact me with [email protected]

Name		Name	Last commit message	Last commit date
Latest commit History 14 Commits
__pycache__		__pycache__
data		data
model		model
README.md		README.md
config.py		config.py
demo.py		demo.py
eval.py		eval.py
evalue.py		evalue.py
export_pb.py		export_pb.py
inference.py		inference.py
train.py		train.py