Skip to content

Latest commit

 

History

History
38 lines (32 loc) · 1.32 KB

File metadata and controls

38 lines (32 loc) · 1.32 KB

pytorch-Pyramid-Attention-Networks-PAN-

train code + inference code were released. [pytorch version]

PAN network:

image

FPA:

image

GAU:

image

How to train

first enter the directory: and follow the instruction: 1.write the dataset path in mypath.py 2. and run

python train.py

after trainning,it will save two model: convnet.pth,pan.pth

How to Inference:

#run:

python inference.py

input:

the Pic in validation dataset:

image

the Pic off the dataset:

image

output:

the Pic in validation dataset:

image

the Pic off the dataset:

image